instruction
stringlengths 999
2.35k
| input
stringlengths 341
562
| output
dict |
|---|---|---|
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['sheets', 'filing_cabinet', 'food_pizza', 'tablelamp', 'blender', 'food_salt', 'blow_dryer', 'trashcan', 'toothbrush', 'cpuscreen', 'hanger', 'ceiling', 'character', 'desk', 'keyboard', 'bookshelf', 'purse', 'ceilinglamp', 'tvstand', 'faucet', 'mousepad', 'microphone', 'oven', 'food_food', 'check', 'tray', 'dresser', 'napkin', 'bathroom_cabinet', 'clothes_shirt']
Task:Leave home
response:
{
"thoughts": Put on shoes. Put on coat. Open door and exit through it. Lock door behind you.,
"actions": ['[WALK] <home_office>', '[WALK] <shoes>', '[FIND] <shoes>', '[GRAB] <shoes>', '[PUTON] <shoes>']
}
|
Scene_Objects:['toilet', 'chair', 'walllamp', 'cpuscreen', 'floor', 'ceiling', 'desk', 'keyboard', 'computer', 'ceilinglamp', 'mat', 'mousepad', 'closetdrawer', 'wall', 'wall_clock', 'powersocket', 'coffe_maker', 'television', 'curtain', 'doorjamb', 'centerpiece', 'mouse', 'shoes', 'dresser', 'door', 'maindoor', 'sink', 'light', 'table', 'soap']
Task:Browse internet
response:
| {
"thoughts": "go to the office, turn computer on, open an browser or any application to open any website.",
"actions": [
"[WALK] <home_office>",
"[WALK] <computer>",
"[FIND] <computer>",
"[SWITCHON] <computer>",
"[FIND] <computer>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['filing_cabinet', 'chair', 'ceilingfan', 'knifeblock', 'orchid', 'clothes_underwear', 'hanger', 'floor', 'ceiling', 'phone', 'character', 'desk', 'computer', 'couch', 'broom', 'wall', 'coffe_maker', 'video_game_controller', 'wine', 'doorjamb', 'towel_rack', 'cards', 'pencil', 'kitchen_counter', 'mouse', 'dresser', 'cupboard', 'door', 'light', 'board_game']
Task:Turn on light
response:
{
"thoughts": Walk over to light switch. Turn on lights.,
"actions": ['[WALK] <light>', '[SWITCHON] <light>']
}
|
Scene_Objects:['chair', 'walllamp', 'detergent', 'cpuscreen', 'hanger', 'floor', 'ceiling', 'desk', 'keyboard', 'computer', 'mail', 'button', 'clothes_jacket', 'ceilinglamp', 'mousepad', 'closetdrawer', 'wall', 'wall_clock', 'powersocket', 'keys', 'television', 'ice', 'doorjamb', 'mouse', 'pencil', 'headset', 'dresser', 'sink', 'light', 'table']
Task:Write an email
response:
| {
"thoughts": "Switch on computer, Use keybord and mouse to open e-mail provider, type requested e-mail, click send, close e-mail provider, turn off computer",
"actions": [
"[WALK] <home_office>",
"[WALK] <computer>",
"[FIND] <computer>",
"[SWITCHON] <computer>",
"[FIND] <mouse>",
"[FIND] <keyboard>",
"[FIND] <mail>",
"[OPEN] <mail>",
"[TYPE] <keyboard>",
"[FIND] <button>",
"[PUSH] <button>",
"[CLOSE] <mail>",
"[SWITCHOFF] <computer>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['filing_cabinet', 'tablelamp', 'chair', 'ground_coffee', 'wallshelf', 'hanger', 'floor', 'stamp', 'desk', 'bookshelf', 'tea_bag', 'envelope', 'computer', 'mop_bucket', 'mat', 'shower', 'foundation', 'form', 'wall', 'powersocket', 'cutting_board', 'photoframe', 'novel', 'check', 'television', 'wooden_spoon', 'after_shave', 'doorjamb', 'door', 'light']
Task:Read to child
response:
{
"thoughts": Read a book out loud to a child,
"actions": ['[WALK] <bedroom>', '[WALK] <child>', '[FIND] <child>', '[GREET] <child>', '[WALK] <filing_cabinet>', '[OPEN] <filing_cabinet>', '[FIND] <novel>', '[GRAB] <novel>', '[WALK] <chair>', '[PULL] <chair>', '[SIT] <chair>', '[TURNTO] <novel>', '[LOOKAT] <novel>', '[TURNTO] <novel>', '[POINTAT] <novel>', '[TURNTO] <novel>', '[POINTAT] <novel>', '[READ] <novel>', '[READ] <novel>', '[READ] <novel>']
}
|
Scene_Objects:['fork', 'filing_cabinet', 'tablelamp', 'mop', 'bathroom', 'drawing', 'floor', 'phone', 'character', 'desk', 'mail', 'laptop', 'tvstand', 'stereo', 'food_orange', 'dining_room', 'wall', 'oven', 'photoframe', 'television', 'tray', 'towel_rack', 'kettle', 'kitchen_counter', 'dresser', 'sink', 'bathtub', 'table', 'board_game', 'coffee']
Task:Walk through
response:
| {
"thoughts": "Walk through the entrance hall",
"actions": [
"[WALK] <home_office>",
"[WALK] <home_office>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['tablelamp', 'food_salt', 'microwave', 'jelly', 'detergent', 'food_carrot', 'toothbrush', 'needle', 'hanger', 'floor', 'phone', 'shampoo', 'character', 'tvstand', 'iron', 'balanceball', 'stereo', 'mousepad', 'food_jam', 'pillow', 'nightstand', 'toothbrush_holder', 'diary', 'console', 'shaving_cream', 'kitchen_counter', 'sink', 'light', 'dishwasher', 'vase']
Task:Listen to music
response:
{
"thoughts": Find the playback device you want to listen to. turn on. press play. enjoyu music.,
"actions": ['[WALK] <bedroom>', '[WALK] <cd_player>', '[FIND] <cd_player>', '[SWITCHON] <cd_player>']
}
|
Scene_Objects:['walllamp', 'bathroom', 'drawing', 'orchid', 'wallshelf', 'bills', 'floor', 'ceiling', 'desk', 'bookshelf', 'remote_control', 'computer', 'stovefan', 'mat', 'tvstand', 'ceilinglamp', 'shower', 'window', 'couch', 'pillow', 'wall', 'powersocket', 'bench', 'photoframe', 'television', 'curtain', 'doorjamb', 'door', 'table', 'light']
Task:Watch TV
response:
| {
"thoughts": "Sit on my couch directly opposite my tv , switch on my tv with the remote control and watch",
"actions": [
"[WALK] <home_office>",
"[WALK] <couch>",
"[FIND] <couch>",
"[SIT] <couch>",
"[FIND] <remote_control>",
"[GRAB] <remote_control>",
"[FIND] <television>",
"[SWITCHON] <television>",
"[PUTOBJBACK] <remote_control>",
"[TURNTO] <television>",
"[WATCH] <television>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['chair', 'walllamp', 'detergent', 'drawing', 'orchid', 'floor', 'character', 'keyboard', 'toaster', 'mat', 'faucet', 'stereo', 'couch', 'towel', 'wall', 'wall_clock', 'nightstand', 'cup', 'coffe_maker', 'kitchen_counter', 'shoes', 'cd', 'hairbrush', 'cupboard', 'sink', 'bathroom_cabinet', 'dishwasher', 'soap', 'scrabble', 'hands_both']
Task:Wash hands
response:
{
"thoughts": I have had my lunch and walked towards the sink and applied soap to my hands and washed cleanly.,
"actions": ['[WALK] <dining_room>', '[WALK] <sink>', '[FIND] <soap>', '[GRAB] <soap>', '[FIND] <hands_both>', '[WASH] <hands_both>', '[FIND] <towel>', '[GRAB] <towel>', '[WIPE] <hands_both>']
}
|
Scene_Objects:['sheets', 'walllamp', 'bathroom', 'trashcan', 'dough', 'drawing', 'wallshelf', 'bathroom_counter', 'floor', 'ceiling', 'bookshelf', 'faucet', 'window', 'food_noodles', 'mousepad', 'pillow', 'dining_room', 'wall', 'bed', 'food_food', 'razor', 'rag', 'console', 'mouse', 'towel_rack', 'kitchen_counter', 'door', 'sink', 'bathtub', 'bathroom_cabinet']
Task:Clean sink
response:
| {
"thoughts": "Take a premoistened cleaning wipe and wipe down top of sink and faucet. Next clean the sink bowl. Let air dry.",
"actions": [
"[WALK] <bathroom>",
"[WALK] <rag>",
"[FIND] <rag>",
"[GRAB] <rag>",
"[WALK] <sink>",
"[FIND] <faucet>",
"[WIPE] <faucet>",
"[WIPE] <sink>",
"[PUTOBJBACK] <rag>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['walllamp', 'nail_polish', 'bathroom', 'detergent', 'knife', 'cpuscreen', 'dvd_player', 'bedroom', 'computer', 'table_cloth', 'mop_bucket', 'brush', 'toaster', 'shower', 'window', 'couch', 'dining_room', 'wall', 'pasta', 'cup', 'photoframe', 'after_shave', 'pencil', 'kitchen_counter', 'dresser', 'clothes_dress', 'sink', 'instrument_guitar', 'light', 'creditcard']
Task:Sit
response:
{
"thoughts": We sit at the dining room table,
"actions": ['[WALK] <dining_room>', '[WALK] <chair>', '[FIND] <chair>', '[PULL] <chair>', '[SIT] <chair>']
}
|
Scene_Objects:['pot', 'tea', 'microwave', 'bathroom', 'trashcan', 'orchid', 'knife', 'drawing', 'clothes_underwear', 'floor', 'ceiling', 'envelope', 'mat', 'mousepad', 'couch', 'pillow', 'wall', 'nightstand', 'novel', 'food_food', 'food_donut', 'curtain', 'doorjamb', 'kitchen_counter', 'hair', 'door', 'sink', 'table', 'light', 'dishwasher']
Task:Put out flowers
response:
| {
"thoughts": "Place a flower centrepiece on the entrance hall table.",
"actions": [
"[WALK] <home_office>",
"[WALK] <centerpiece>",
"[FIND] <centerpiece>",
"[GRAB] <centerpiece>",
"[WALK] <table>",
"[PUTBACK] <centerpiece> <table>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['walllamp', 'food_carrot', 'drawing', 'orchid', 'food_egg', 'bathroom_counter', 'floor', 'phone', 'character', 'desk', 'bookshelf', 'stovefan', 'ceilinglamp', 'shower', 'faucet', 'window', 'oil', 'couch', 'clothes_pants', 'pillow', 'wall', 'tape', 'cup', 'novel', 'tray', 'curtain', 'mouse', 'chessboard', 'controller', 'table']
Task:Read book
response:
{
"thoughts": Sit on a couch, grab a book, open it, and scan the pages,
"actions": ['[WALK] <home_office>', '[WALK] <novel>', '[FIND] <novel>', '[GRAB] <novel>', '[WALK] <couch>', '[SIT] <couch>', '[TURNTO] <novel>', '[LOOKAT] <novel>', '[READ] <novel>', '[READ] <novel>', '[READ] <novel>']
}
|
Scene_Objects:['food_pizza', 'tablelamp', 'toilet', 'chair', 'detergent', 'trashcan', 'orchid', 'floor', 'ceiling', 'stamp', 'man', 'bookshelf', 'purse', 'toaster', 'mat', 'oil', 'mousepad', 'nightstand', 'cup', 'photoframe', 'newspaper', 'coffe_maker', 'tray', 'curtain', 'mouse', 'headset', 'cupboard', 'door', 'bathroom_cabinet', 'dishwasher']
Task:Answer door
response:
| {
"thoughts": "Wait until doorbell rings, look to see who it is outside and open the door if you feel so inclined.",
"actions": [
"[WALK] <home_office>",
"[WALK] <man>",
"[FIND] <man>",
"[GREET] <man>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['chair', 'freezer', 'walllamp', 'orchid', 'cpuscreen', 'floor', 'ceiling', 'desk', 'keyboard', 'computer', 'toaster', 'ceilinglamp', 'tvstand', 'faucet', 'stereo', 'window', 'shower', 'mousepad', 'wall', 'powersocket', 'bench', 'clothes_gloves', 'television', 'tray', 'rag', 'doorjamb', 'towel_rack', 'mouse', 'bathtub', 'table']
Task:Work
response:
{
"thoughts": I walk into home office. I walk to desk. I sit at chair. I turn on computer.,
"actions": ['[WALK] <home_office>', '[WALK] <desk>', '[FIND] <chair>', '[SIT] <chair>', '[FIND] <computer>', '[SWITCHON] <computer>']
}
|
Scene_Objects:['sponge', 'pot', 'microwave', 'chair', 'walllamp', 'bathroom', 'knifeblock', 'needle', 'orchid', 'cpuscreen', 'floor', 'ceiling', 'character', 'desk', 'plate', 'faucet', 'dish_soap', 'pillow', 'wall', 'oven', 'box', 'coffe_maker', 'coin', 'mouse', 'kitchen_counter', 'dishrack', 'cupboard', 'sink', 'home_office', 'table']
Task:Wash dishes
response:
| {
"thoughts": "Turn water on and run over dishes while scrubbing with a soapy sponge.",
"actions": [
"[WALK] <dining_room>",
"[WALK] <sink>",
"[FIND] <faucet>",
"[SWITCHON] <faucet>",
"[FIND] <dish_soap>",
"[GRAB] <dish_soap>",
"[FIND] <sponge>",
"[GRAB] <sponge>",
"[POUR] <dish_soap> <sponge>",
"[PUTOBJBACK] <dish_soap>",
"[FIND] <plate>",
"[GRAB] <plate>",
"[SCRUB] <plate>",
"[RINSE] <plate>",
"[FIND] <dishrack>",
"[PUTBACK] <plate> <dishrack>",
"[FIND] <plate>",
"[GRAB] <plate>",
"[SCRUB] <plate>",
"[RINSE] <plate>",
"[PUTBACK] <plate> <dishrack>",
"[RINSE] <sponge>",
"[PUTOBJBACK] <sponge>",
"[SWITCHOFF] <faucet>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['sponge', 'filing_cabinet', 'juice', 'microwave', 'knifeblock', 'wallshelf', 'bathroom_counter', 'floor', 'ceiling', 'computer', 'piano_bench', 'toaster', 'ceilinglamp', 'plate', 'faucet', 'stovefan', 'tvstand', 'dish_soap', 'wall', 'pasta', 'coffee_filter', 'tray', 'doorjamb', 'mouse', 'kitchen_counter', 'shoes', 'dishrack', 'cupboard', 'sink', 'scrabble']
Task:Wash dishes
response:
{
"thoughts": Place all of the dirty dishes in a sink full of clean, soapy water. Wash them. Rinse them in fresh, running water. Place them in the strainer to dry.,
"actions": ['[WALK] <dining_room>', '[WALK] <sink>', '[FIND] <faucet>', '[SWITCHON] <faucet>', '[FIND] <plate>', '[GRAB] <plate>', '[RINSE] <plate>', '[PUTBACK] <plate> <sink>', '[FIND] <plate>', '[GRAB] <plate>', '[RINSE] <plate>', '[PUTBACK] <plate> <sink>', '[SWITCHOFF] <faucet>', '[FIND] <dish_soap>', '[GRAB] <dish_soap>', '[FIND] <sponge>', '[GRAB] <sponge>', '[POUR] <dish_soap> <sponge>', '[PUTOBJBACK] <dish_soap>', '[GRAB] <plate>', '[SCRUB] <plate>', '[PUTOBJBACK] <plate>', '[GRAB] <plate>', '[SCRUB] <plate>', '[PUTOBJBACK] <plate>', '[SWITCHON] <faucet>', '[RINSE] <sponge>', '[PUTOBJBACK] <sponge>', '[GRAB] <plate>', '[RINSE] <plate>', '[FIND] <dishrack>', '[PUTBACK] <plate> <dishrack>', '[GRAB] <plate>', '[RINSE] <plate>', '[PUTBACK] <plate> <dishrack>', '[SWITCHOFF] <faucet>']
}
|
Scene_Objects:['spectacles', 'filing_cabinet', 'food_salt', 'mop', 'microwave', 'band-aids', 'freezer', 'walllamp', 'bathroom', 'food_carrot', 'drawing', 'floor', 'ceiling', 'keyboard', 'bookshelf', 'computer', 'faucet', 'mousepad', 'couch', 'pillow', 'wall', 'photoframe', 'tray', 'curtain', 'doorjamb', 'shoes', 'door', 'bathroom_cabinet', 'light', 'shoe_rack']
Task:Turn light off
response:
| {
"thoughts": "Put the light switch into the down, or off, position.",
"actions": [
"[WALK] <bedroom>",
"[WALK] <light>",
"[SWITCHOFF] <light>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['tablelamp', 'juice', 'measuring_cup', 'chair', 'trashcan', 'ceilingfan', 'drawing', 'cpuscreen', 'floor', 'ceiling', 'phone', 'desk', 'keyboard', 'bookshelf', 'computer', 'ceilinglamp', 'balanceball', 'mousepad', 'wall', 'nightstand', 'doorjamb', 'mouse', 'pencil', 'cupboard', 'door', 'maindoor', 'home_office', 'light', 'dishwasher', 'creditcard']
Task:Work
response:
{
"thoughts": turn computer on, open application to work with,
"actions": ['[WALK] <home_office>', '[WALK] <computer>', '[FIND] <computer>', '[SWITCHON] <computer>']
}
|
Scene_Objects:['sheets', 'pot', 'filing_cabinet', 'toilet', 'band-aids', 'knifeblock', 'orchid', 'bedroom', 'keyboard', 'thread', 'clothes_jacket', 'faucet', 'food_noodles', 'food_orange', 'pillow', 'dining_room', 'oven', 'bench', 'powersocket', 'cup', 'newspaper', 'bed', 'check', 'coffe_maker', 'ice', 'tray', 'towel_rack', 'bathtub', 'home_office', 'dishwasher']
Task:Walk through
response:
| {
"thoughts": "Walk through the hallway.",
"actions": [
"[WALK] <home_office>",
"[WALK] <home_office>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['pot', 'mop', 'toilet', 'chair', 'freezer', 'ceilingfan', 'drawing', 'knifeblock', 'cpuscreen', 'floor', 'ceiling', 'remote_control', 'computer', 'button', 'ceilinglamp', 'tvstand', 'balanceball', 'window', 'couch', 'wall', 'wall_clock', 'bench', 'television', 'wine', 'curtain', 'coffee_pot', 'door', 'sink', 'spoon', 'table']
Task:Watch TV
response:
{
"thoughts": I watch Netflix from about 8:30 to 10:30 every night. I usually watch TV shows.,
"actions": ['[WALK] <home_office>', '[WALK] <wall_clock>', '[FIND] <wall_clock>', '[TURNTO] <wall_clock>', '[LOOKAT] <wall_clock>', '[FIND] <couch>', '[SIT] <couch>', '[FIND] <remote_control>', '[GRAB] <remote_control>', '[FIND] <television>', '[SWITCHON] <television>', '[PUTOBJBACK] <remote_control>', '[TURNTO] <television>', '[WATCH] <television>']
}
|
Scene_Objects:['fork', 'pot', 'microwave', 'walllamp', 'bathroom', 'detergent', 'knifeblock', 'knife', 'clothes_underwear', 'kitchen_cabinet', 'floor', 'ceiling', 'phone', 'toaster', 'stovefan', 'plate', 'faucet', 'tvstand', 'wall', 'nightstand', 'oven', 'coffe_maker', 'drinking_glass', 'tray', 'kitchen_counter', 'cupboard', 'sink', 'spoon', 'home_office', 'dishwasher']
Task:Unload dishwasher
response:
| {
"thoughts": "Open the dishwasher door. Take out all of the dishes and silverware. Put the dishes and silverware in their drawers and cabinets. Close the dishwasher door.",
"actions": [
"[WALK] <dining_room>",
"[WALK] <dishwasher>",
"[FIND] <dishwasher>",
"[OPEN] <dishwasher>",
"[FIND] <drinking_glass>",
"[GRAB] <drinking_glass>",
"[FIND] <cupboard>",
"[OPEN] <cupboard>",
"[PUTBACK] <drinking_glass> <cupboard>",
"[FIND] <fork>",
"[GRAB] <fork>",
"[FIND] <kitchen_cabinet>",
"[OPEN] <kitchen_cabinet>",
"[PUTIN] <fork> <kitchen_cabinet>",
"[FIND] <knife>",
"[GRAB] <knife>",
"[PUTIN] <knife> <kitchen_cabinet>",
"[FIND] <spoon>",
"[GRAB] <spoon>",
"[PUTIN] <spoon> <kitchen_cabinet>",
"[FIND] <plate>",
"[GRAB] <plate>",
"[PUTIN] <plate> <kitchen_cabinet>",
"[CLOSE] <kitchen_cabinet>",
"[CLOSE] <dishwasher>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['walllamp', 'drawing', 'hanger', 'bathroom_counter', 'floor', 'ceiling', 'folder', 'bedroom', 'bookshelf', 'computer', 'ceilinglamp', 'deck_of_cards', 'pillow', 'dining_room', 'wall', 'bench', 'toy', 'box', 'photoframe', 'clothes_gloves', 'novel', 'video_game_controller', 'cards', 'vacuum_cleaner', 'dresser', 'controller', 'maindoor', 'napkin', 'dishwasher', 'instrument_violin']
Task:Pick up toys
response:
{
"thoughts": Bend down and pick up the toys off of the floor.,
"actions": ['[WALK] <bedroom>', '[WALK] <toy>', '[FIND] <toy>', '[GRAB] <toy>', '[WALK] <dresser>', '[OPEN] <dresser>', '[PUTIN] <toy> <dresser>', '[FIND] <toy>', '[GRAB] <toy>', '[WALK] <dresser>', '[PUTIN] <toy> <dresser>', '[FIND] <toy>', '[WALK] <dresser>', '[GRAB] <toy>', '[PUTIN] <toy> <dresser>', '[CLOSE] <dresser>']
}
|
Scene_Objects:['pot', 'microwave', 'chair', 'freezer', 'walllamp', 'detergent', 'milk', 'knifeblock', 'drawing', 'cpuscreen', 'kitchen_cabinet', 'floor', 'ceiling', 'computer', 'stovefan', 'toaster', 'tvstand', 'faucet', 'iron', 'couch', 'wall', 'oven', 'food_food', 'television', 'tray', 'kitchen_counter', 'cupboard', 'sink', 'laser_pointer', 'scrabble']
Task:Cook some food
response:
| {
"thoughts": "Cooking a meal",
"actions": [
"[WALK] <dining_room>",
"[WALK] <kitchen_cabinet>",
"[FIND] <kitchen_cabinet>",
"[OPEN] <kitchen_cabinet>",
"[FIND] <pot>",
"[GRAB] <pot>",
"[CLOSE] <kitchen_cabinet>",
"[WALK] <sink>",
"[FIND] <faucet>",
"[SWITCHON] <faucet>",
"[PUTBACK] <pot> <sink>",
"[GRAB] <pot>",
"[SWITCHOFF] <faucet>",
"[WALK] <oven>",
"[PUTBACK] <pot> <oven>",
"[WALK] <freezer>",
"[OPEN] <freezer>",
"[FIND] <food_food>",
"[GRAB] <food_food>",
"[CLOSE] <freezer>",
"[WALK] <oven>",
"[PUTBACK] <food_food> <pot>",
"[SWITCHON] <oven>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['tablelamp', 'chair', 'walllamp', 'drawing', 'orchid', 'cpuscreen', 'wallshelf', 'hanger', 'floor', 'ceiling', 'phone', 'desk', 'keyboard', 'computer', 'mat', 'ceilinglamp', 'tvstand', 'faucet', 'mousepad', 'couch', 'dustpan', 'pillow', 'wall', 'nightstand', 'television', 'doorjamb', 'mouse', 'dresser', 'cupboard', 'door']
Task:Watch youtube
response:
{
"thoughts": Go to youtube.Com watch videos,
"actions": ['[WALK] <home_office>', '[WALK] <chair>', '[FIND] <chair>', '[SIT] <chair>', '[FIND] <computer>', '[SWITCHON] <computer>', '[TURNTO] <computer>', '[LOOKAT] <computer>']
}
|
Scene_Objects:['microwave', 'toilet', 'dog', 'freezer', 'walllamp', 'bathroom', 'trashcan', 'drawing', 'knifeblock', 'cpuscreen', 'phone', 'bookshelf', 'table_cloth', 'bookmark', 'toaster', 'mat', 'faucet', 'oven', 'cup', 'photoframe', 'check', 'television', 'towel_rack', 'kitchen_counter', 'chessboard', 'cupboard', 'door', 'sink', 'light', 'tooth_paste']
Task:Pet dog
response:
| {
"thoughts": "Pet my dog in hall",
"actions": [
"[WALK] <home_office>",
"[WALK] <dog>",
"[TURNTO] <dog>",
"[LOOKAT] <dog>",
"[TOUCH] <dog>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['tablelamp', 'food_ice_cream', 'freezer', 'cat', 'knifeblock', 'longboard', 'floor', 'ceiling', 'keyboard', 'clothes_jacket', 'stovefan', 'ceilinglamp', 'mat', 'stereo', 'tvstand', 'electric_shaver', 'mousepad', 'couch', 'wall', 'oven', 'bench', 'cup', 'food_food', 'doorjamb', 'conditioner', 'clothes_scarf', 'home_office', 'table', 'light', 'vase']
Task:Put groceries in Fridge
response:
{
"thoughts": I get my groceries and go to the fridge. I open the fridge and put the groceries there.,
"actions": ['[WALK] <dining_room>', '[WALK] <food_food>', '[FIND] <food_food>', '[GRAB] <food_food>', '[FIND] <freezer>', '[OPEN] <freezer>', '[PUTIN] <food_food> <freezer>', '[CLOSE] <freezer>']
}
|
Scene_Objects:['sponge', 'pot', 'bathroom', 'knifeblock', 'bathroom_counter', 'shampoo', 'phone', 'mail', 'clothes_jacket', 'stovefan', 'ceilinglamp', 'window', 'clothes_socks', 'couch', 'clothes_pants', 'wall', 'powersocket', 'oven', 'photoframe', 'coffe_maker', 'television', 'mouse', 'shoes', 'chessboard', 'dresser', 'cupboard', 'bathroom_cabinet', 'table', 'dishwasher', 'clothes_shirt']
Task:Get dressed
response:
| {
"thoughts": "Open drawer. Pull out pants and put them on. Take a shirt out and put it on. Put on socks and shoes.",
"actions": [
"[WALK] <bedroom>",
"[WALK] <dresser>",
"[OPEN] <dresser>",
"[FIND] <clothes_pants>",
"[GRAB] <clothes_pants>",
"[PUTON] <clothes_pants>",
"[FIND] <clothes_shirt>",
"[GRAB] <clothes_shirt>",
"[PUTON] <clothes_shirt>",
"[FIND] <clothes_socks>",
"[GRAB] <clothes_socks>",
"[PUTON] <clothes_socks>",
"[CLOSE] <dresser>",
"[FIND] <shoes>",
"[GRAB] <shoes>",
"[PUTON] <shoes>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['tablelamp', 'chair', 'walllamp', 'drawing', 'orchid', 'cpuscreen', 'wallshelf', 'hanger', 'floor', 'ceiling', 'phone', 'desk', 'keyboard', 'computer', 'mat', 'ceilinglamp', 'tvstand', 'mousepad', 'couch', 'pillow', 'wall', 'nightstand', 'oven', 'cup', 'television', 'tray', 'doorjamb', 'mouse', 'dresser', 'door']
Task:Watch youtube
response:
{
"thoughts": Go to youtube.Com watch videos,
"actions": ['[WALK] <home_office>', '[WALK] <chair>', '[FIND] <chair>', '[SIT] <chair>', '[FIND] <computer>', '[SWITCHON] <computer>', '[FIND] <keyboard>', '[TYPE] <keyboard>', '[TURNTO] <computer>', '[LOOKAT] <computer>']
}
|
Scene_Objects:['chair', 'walllamp', 'bathroom', 'trashcan', 'orchid', 'drawing', 'cpuscreen', 'floor', 'ceiling', 'folder', 'bedroom', 'bookshelf', 'keyboard', 'ceilinglamp', 'faucet', 'window', 'couch', 'pillow', 'wall', 'powersocket', 'food_food', 'television', 'curtain', 'doorjamb', 'towel_rack', 'dresser', 'bathtub', 'light', 'table', 'instrument_violin']
Task:Straighten pictures on wall
response:
| {
"thoughts": "Move the pictures/paintings until they are straight.",
"actions": [
"[WALK] <home_office>",
"[WALK] <drawing>",
"[FIND] <drawing>",
"[TURNTO] <drawing>",
"[LOOKAT] <drawing>",
"[TOUCH] <drawing>",
"[PUSH] <drawing>",
"[FIND] <drawing>",
"[TURNTO] <drawing>",
"[LOOKAT] <drawing>",
"[TOUCH] <drawing>",
"[PUSH] <drawing>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['sheets', 'spectacles', 'filing_cabinet', 'food_salt', 'walllamp', 'drawing', 'bathroom_counter', 'floor', 'character', 'shower', 'stereo', 'microphone', 'wall', 'toy', 'cutting_board', 'cup', 'novel', 'food_food', 'check', 'food_onion', 'coffee_pot', 'console', 'shaving_cream', 'towel_rack', 'shoes', 'cd', 'hairbrush', 'maindoor', 'soap', 'coffee']
Task:Read
response:
{
"thoughts": Read a book in the living room,
"actions": ['[WALK] <home_office>', '[WALK] <filing_cabinet>', '[OPEN] <filing_cabinet>', '[FIND] <novel>', '[GRAB] <novel>', '[CLOSE] <filing_cabinet>', '[READ] <novel>']
}
|
Scene_Objects:['pot', 'food_pizza', 'tea', 'toilet', 'walllamp', 'nail_polish', 'cpuscreen', 'floor', 'phone', 'ceiling', 'keyboard', 'toaster', 'mat', 'oil', 'pillow', 'bowl', 'wall', 'powersocket', 'oven', 'photoframe', 'television', 'rag', 'doorjamb', 'headset', 'door', 'sink', 'home_office', 'light', 'bathroom_cabinet', 'coffee']
Task:Turn light off
response:
| {
"thoughts": "Put the light switch into the down, or off, position.",
"actions": [
"[WALK] <dining_room>",
"[WALK] <light>",
"[FIND] <light>",
"[SWITCHOFF] <light>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['toilet', 'chair', 'freezer', 'detergent', 'hanger', 'bathroom_counter', 'shampoo', 'ceiling', 'desk', 'bedroom', 'computer', 'stovefan', 'toaster', 'tvstand', 'shower', 'homework', 'closetdrawer', 'bench', 'oven', 'toy', 'cup', 'bed', 'photoframe', 'toilet_paper', 'coffe_maker', 'oven_mitts', 'pencil', 'dresser', 'bag', 'bathroom_cabinet']
Task:Pet cat
response:
{
"thoughts": I enter the room where the cat is. I look for the cat. I find the cat. I walk over to the cat. I use my hand to pet the cat.,
"actions": ['[WALK] <home_office>', '[WALK] <cat>', '[FIND] <cat>', '[TURNTO] <cat>', '[LOOKAT] <cat>', '[FIND] <cat>', '[WALK] <cat>', '[TOUCH] <cat>']
}
|
Scene_Objects:['chair', 'trashcan', 'orchid', 'drawing', 'floor', 'ceiling', 'bookshelf', 'computer', 'stovefan', 'ceilinglamp', 'mat', 'shower', 'window', 'food_noodles', 'couch', 'pillow', 'wall', 'photoframe', 'toothbrush_holder', 'curtain', 'doorjamb', 'chessboard', 'door', 'bathtub', 'bag', 'home_office', 'light', 'table', 'dishwasher', 'instrument_violin']
Task:Admire art
response:
| {
"thoughts": "Looking at painting, art",
"actions": [
"[WALK] <home_office>",
"[WALK] <drawing>",
"[TURNTO] <drawing>",
"[LOOKAT] <drawing>",
"[FIND] <drawing>",
"[TURNTO] <drawing>",
"[LOOKAT] <drawing>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['sheets', 'microwave', 'chair', 'cat', 'dough', 'bathroom_counter', 'phone', 'bookshelf', 'ceilinglamp', 'tvstand', 'stereo', 'candle', 'bowl', 'dining_room', 'wall', 'newspaper', 'coffe_maker', 'food_food', 'coffee_filter', 'tray', 'curtain', 'diary', 'towel_rack', 'pencil', 'dresser', 'food_steak', 'sink', 'napkin', 'door', 'dishwasher']
Task:Pick up phone
response:
{
"thoughts": I'll enter the location containing the phone. Once there, I'll locate the phone, and move close to it. I'll then use my hand to pick up the phone such that I can then use it.,
"actions": ['[WALK] <home_office>', '[WALK] <phone>', '[FIND] <phone>', '[WALK] <phone>', '[TURNTO] <phone>', '[LOOKAT] <phone>', '[GRAB] <phone>']
}
|
Scene_Objects:['microwave', 'walllamp', 'trashcan', 'orchid', 'drawing', 'wallshelf', 'bathroom_counter', 'floor', 'ceiling', 'phone', 'bookshelf', 'remote_control', 'mat', 'tvstand', 'ceilinglamp', 'shower', 'window', 'couch', 'pillow', 'wall', 'nightstand', 'powersocket', 'photoframe', 'television', 'curtain', 'doorjamb', 'kitchen_counter', 'door', 'light', 'table']
Task:Watch TV
response:
| {
"thoughts": "I sit down on the couch, get the tv remote, click the on button, and change the channel to the show i want to watch",
"actions": [
"[WALK] <home_office>",
"[WALK] <television>",
"[FIND] <television>",
"[SWITCHON] <television>",
"[WALK] <couch>",
"[SIT] <couch>",
"[FIND] <remote_control>",
"[GRAB] <remote_control>",
"[TURNTO] <television>",
"[POINTAT] <television>",
"[PUTOBJBACK] <remote_control>",
"[WATCH] <television>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['filing_cabinet', 'cd_player', 'blow_dryer', 'bathroom', 'detergent', 'orchid', 'bills', 'bathroom_counter', 'phone', 'bedroom', 'remote_control', 'thread', 'button', 'mat', 'ceilinglamp', 'couch', 'pillow', 'dining_room', 'broom', 'wall_clock', 'powersocket', 'bench', 'box', 'coffe_maker', 'razor', 'towel_rack', 'vacuum_cleaner', 'bathtub', 'dishwasher', 'soap']
Task:Listen to music
response:
{
"thoughts": I walk to my living room. Find CD player. Plugin the CD player to Power. Also find the remote control. Push the Botton,
"actions": ['[WALK] <home_office>', '[WALK] <cd_player>', '[FIND] <cd_player>', '[PLUGIN] <cd_player>', '[FIND] <remote_control>', '[GRAB] <remote_control>', '[FIND] <button>', '[PUSH] <button>']
}
|
Scene_Objects:['toilet', 'walllamp', 'bathroom', 'drawing', 'orchid', 'wallshelf', 'cpuscreen', 'floor', 'ceiling', 'character', 'bookshelf', 'mat', 'tvstand', 'ceilinglamp', 'shower', 'window', 'couch', 'pillow', 'wall', 'powersocket', 'photoframe', 'bed', 'television', 'curtain', 'doorjamb', 'shoes', 'door', 'table', 'light', 'creditcard']
Task:Watch TV
response:
| {
"thoughts": "Sit on my couch directly opposite my tv , switch on my tv with the remote control and watch",
"actions": [
"[WALK] <home_office>",
"[WALK] <remote_control>",
"[FIND] <remote_control>",
"[GRAB] <remote_control>",
"[WALK] <couch>",
"[SIT] <couch>",
"[TOUCH] <remote_control>",
"[FIND] <television>",
"[SWITCHON] <television>",
"[TURNTO] <television>",
"[WATCH] <television>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['spectacles', 'glue', 'filing_cabinet', 'microwave', 'jelly', 'freezer', 'ceilingfan', 'orchid', 'food_egg', 'floor', 'phone', 'ceiling', 'desk', 'thread', 'mat', 'laptop', 'deck_of_cards', 'faucet', 'foundation', 'pillow', 'nightstand', 'powersocket', 'basket_for_clothes', 'food_food', 'lighter', 'sink', 'home_office', 'vase', 'soap', 'washing_machine']
Task:Wash clothes
response:
{
"thoughts": I go to the bedroom to find and grab basket of clothes. I then go to the bathroom and find the washing machine. I put clothes inside the washing machine and put washing powder. I switch on the washing machine.,
"actions": ['[WALK] <bedroom>', '[WALK] <basket_for_clothes>', '[FIND] <basket_for_clothes>', '[GRAB] <basket_for_clothes>', '[WALK] <bathroom>', '[WALK] <washing_machine>', '[FIND] <washing_machine>', '[OPEN] <washing_machine>', '[PUTBACK] <basket_for_clothes> <washing_machine>', '[FIND] <soap>', '[GRAB] <soap>', '[PUTBACK] <soap> <washing_machine>', '[FIND] <washing_machine>', '[CLOSE] <washing_machine>', '[SWITCHON] <washing_machine>']
}
|
Scene_Objects:['tablelamp', 'bathroom', 'trashcan', 'orchid', 'drawing', 'wallshelf', 'floor', 'character', 'bedroom', 'tea_bag', 'keyboard', 'toaster', 'mat', 'stovefan', 'tvstand', 'couch', 'dining_room', 'wall', 'nightstand', 'child', 'bed', 'photoframe', 'novel', 'food_onion', 'after_shave', 'cupboard', 'bathtub', 'napkin', 'bathroom_cabinet', 'board_game']
Task:Story reading time
response:
| {
"thoughts": "Sit on bed with child, open mother goose, pick story, read book, point at the pretty pictures",
"actions": [
"[WALK] <bedroom>",
"[WALK] <child>",
"[FIND] <child>",
"[GREET] <child>",
"[FIND] <novel>",
"[GRAB] <novel>",
"[FIND] <bed>",
"[SIT] <bed>",
"[TURNTO] <child>",
"[LOOKAT] <child>",
"[TURNTO] <novel>",
"[POINTAT] <novel>",
"[READ] <novel>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['pot', 'scissors', 'ceilingfan', 'drawing', 'hanger', 'longboard', 'floor', 'ceiling', 'remote_control', 'ceilinglamp', 'tvstand', 'balanceball', 'shower', 'window', 'couch', 'food_kiwi', 'wall', 'oven', 'powersocket', 'cup', 'light_bulb', 'novel', 'television', 'tray', 'curtain', 'after_shave', 'kitchen_counter', 'dresser', 'table', 'light']
Task:Watch TV
response:
{
"thoughts": walk to living room, find couch, sit on couch, find remote control , turn on tv by pressing button,
"actions": ['[WALK] <home_office>', '[WALK] <couch>', '[FIND] <couch>', '[SIT] <couch>', '[FIND] <remote_control>', '[FIND] <television>', '[TURNTO] <television>']
}
|
Scene_Objects:['tablelamp', 'measuring_cup', 'walllamp', 'drawing', 'orchid', 'wallshelf', 'bathroom_counter', 'floor', 'ceiling', 'bookshelf', 'remote_control', 'keyboard', 'mat', 'tvstand', 'ceilinglamp', 'shower', 'window', 'couch', 'pillow', 'wall', 'powersocket', 'photoframe', 'television', 'curtain', 'doorjamb', 'centerpiece', 'shoes', 'door', 'table', 'light']
Task:Watch TV
response:
| {
"thoughts": "TURN ON TV, SIT ON COUCH ACROSS THE ROOM, USE REMOTE TO LOOK FOR SHOW TO WATCH",
"actions": [
"[WALK] <home_office>",
"[WALK] <couch>",
"[FIND] <couch>",
"[SIT] <couch>",
"[FIND] <remote_control>",
"[GRAB] <remote_control>",
"[FIND] <television>",
"[SWITCHON] <television>",
"[TURNTO] <television>",
"[WATCH] <television>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['measuring_cup', 'chair', 'walllamp', 'bathroom', 'milk', 'orchid', 'cpuscreen', 'floor', 'ceiling', 'desk', 'bedroom', 'keyboard', 'envelope', 'computer', 'mat', 'ceilinglamp', 'stereo', 'mousepad', 'wall', 'toothbrush_holder', 'food_rice', 'ice', 'after_shave', 'mouse', 'headset', 'dresser', 'door', 'home_office', 'table', 'board_game']
Task:Work
response:
{
"thoughts": Walk to home office, walk to desk, sit in chair, turn on computer, look at computer, click mouse and type on keyboard.,
"actions": ['[WALK] <home_office>', '[WALK] <desk>', '[FIND] <chair>', '[SIT] <chair>', '[FIND] <computer>', '[SWITCHON] <computer>', '[TURNTO] <computer>', '[LOOKAT] <computer>', '[FIND] <mouse>', '[TOUCH] <mouse>', '[FIND] <keyboard>', '[TYPE] <keyboard>', '[TURNTO] <computer>', '[LOOKAT] <computer>']
}
|
Scene_Objects:['pot', 'toilet', 'chair', 'freezer', 'walllamp', 'dough', 'woman', 'knifeblock', 'bathroom_counter', 'floor', 'character', 'desk', 'tea_bag', 'bookshelf', 'computer', 'stovefan', 'ceilinglamp', 'wall', 'child', 'coffe_maker', 'video_game_controller', 'television', 'towel_rack', 'kettle', 'kitchen_counter', 'dresser', 'door', 'bathtub', 'instrument_violin', 'clothes_shirt']
Task:Greet guests
response:
| {
"thoughts": "Go to the kitchen, greet your family members",
"actions": [
"[WALK] <dining_room>",
"[WALK] <child>",
"[FIND] <child>",
"[GREET] <child>",
"[FIND] <woman>",
"[GREET] <woman>",
"[FIND] <chair>",
"[SIT] <chair>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['microwave', 'drawing', 'knife', 'hanger', 'floor', 'ceiling', 'bedroom', 'computer', 'alcohol', 'mat', 'ceilinglamp', 'faucet', 'window', 'closetdrawer', 'couch', 'pillow', 'wall', 'bench', 'cutting_board', 'standingmirror', 'food_food', 'television', 'tray', 'curtain', 'doorjamb', 'mouse', 'chessboard', 'dresser', 'table', 'clothes_skirt']
Task:Watch TV
response:
{
"thoughts": Walk into the living room. Sit on the couch. Turn on TV.,
"actions": ['[WALK] <home_office>', '[WALK] <television>', '[FIND] <television>', '[SWITCHON] <television>']
}
|
Scene_Objects:['sheets', 'pot', 'mop', 'walllamp', 'orchid', 'wallshelf', 'floor', 'ceiling', 'phone', 'bedroom', 'bookshelf', 'remote_control', 'piano_bench', 'mat', 'tvstand', 'ceilinglamp', 'shower', 'window', 'wall', 'powersocket', 'box', 'love_seat', 'photoframe', 'television', 'curtain', 'doorjamb', 'dresser', 'clothes_dress', 'bathtub', 'light']
Task:Watch TV
response:
| {
"thoughts": "I'll watch TV",
"actions": [
"[WALK] <home_office>",
"[WALK] <remote_control>",
"[FIND] <remote_control>",
"[GRAB] <remote_control>",
"[FIND] <television>",
"[SWITCHON] <television>",
"[PUTOBJBACK] <remote_control>",
"[FIND] <love_seat>",
"[SIT] <love_seat>",
"[TURNTO] <television>",
"[WATCH] <television>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['fork', 'pot', 'filing_cabinet', 'tablelamp', 'microwave', 'chair', 'bathroom', 'drawing', 'cpuscreen', 'floor', 'phone', 'desk', 'bedroom', 'ceilinglamp', 'shoe-shine_kit', 'faucet', 'clothes_pants', 'dining_room', 'wall', 'tape', 'basket_for_clothes', 'bed', 'coffe_maker', 'chessboard', 'clothes_dress', 'door', 'home_office', 'light', 'soap', 'washing_machine']
Task:Wash clothes
response:
{
"thoughts": Walk to Bathroom. Find basket for clothes. Lift clothes dress from basket. Find washing machine. Walk to washing machine. Put clothes in washing machine. Put soap in washing machine. Switch on washing machine.,
"actions": ['[WALK] <bathroom>', '[WALK] <basket_for_clothes>', '[FIND] <basket_for_clothes>', '[FIND] <clothes_dress>', '[GRAB] <clothes_dress>', '[FIND] <washing_machine>', '[WALK] <washing_machine>', '[OPEN] <washing_machine>', '[PUTBACK] <clothes_dress> <washing_machine>', '[FIND] <soap>', '[GRAB] <soap>', '[PUTBACK] <soap> <washing_machine>', '[CLOSE] <washing_machine>', '[SWITCHON] <washing_machine>']
}
|
Scene_Objects:['filing_cabinet', 'tablelamp', 'microwave', 'walllamp', 'bathroom', 'detergent', 'trashcan', 'drawing', 'bathroom_counter', 'floor', 'ceiling', 'bookshelf', 'mat', 'faucet', 'wall', 'cup', 'photoframe', 'keys', 'ice', 'tray', 'rag', 'towel_rack', 'cupboard', 'sink', 'bathtub', 'bathroom_cabinet', 'home_office', 'dishwasher', 'table', 'laser_pointer']
Task:Wipe down sink
response:
| {
"thoughts": "Use a cloth to wipe the edge of the sink. Then wipe the bowl of the sink as well.",
"actions": [
"[WALK] <bathroom>",
"[WALK] <bathroom_cabinet>",
"[OPEN] <bathroom_cabinet>",
"[FIND] <rag>",
"[GRAB] <rag>",
"[WALK] <sink>",
"[WIPE] <sink>",
"[PUTOBJBACK] <rag>",
"[CLOSE] <bathroom_cabinet>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['measuring_cup', 'food_carrot', 'drawing', 'floor', 'thread', 'toaster', 'mat', 'picture', 'oil', 'towel', 'nightstand', 'powersocket', 'toy', 'shelf', 'box', 'cup', 'cutting_board', 'bed', 'photoframe', 'novel', 'toilet_paper', 'check', 'television', 'tray', 'food_onion', 'curtain', 'dresser', 'door', 'hairdryer', 'tooth_paste']
Task:Listen to music
response:
{
"thoughts": Find the radio; switch on the radio and listen to music.,
"actions": ['[WALK] <home_office>', '[WALK] <stereo>', '[FIND] <stereo>', '[SWITCHON] <stereo>']
}
|
Scene_Objects:['walllamp', 'drawing', 'orchid', 'wallshelf', 'cpuscreen', 'floor', 'ceiling', 'desk', 'bookshelf', 'remote_control', 'mat', 'tvstand', 'ceilinglamp', 'shower', 'window', 'couch', 'dustpan', 'pillow', 'wall', 'nightstand', 'powersocket', 'photoframe', 'television', 'curtain', 'doorjamb', 'towel_rack', 'pencil', 'door', 'table', 'light']
Task:Watch TV
response:
| {
"thoughts": "In the evenings after dinner, we sit on the couch, turn on the television, cue up Netflix and watch our favorite shows.",
"actions": [
"[WALK] <home_office>",
"[WALK] <couch>",
"[FIND] <couch>",
"[FIND] <remote_control>",
"[GRAB] <remote_control>",
"[FIND] <television>",
"[SWITCHON] <television>",
"[SIT] <couch>",
"[TURNTO] <television>",
"[WATCH] <television>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['filing_cabinet', 'tablelamp', 'beer', 'chair', 'trashcan', 'ceilingfan', 'drawing', 'orchid', 'cpuscreen', 'floor', 'ceiling', 'phone', 'desk', 'keyboard', 'bookshelf', 'computer', 'table_cloth', 'stovefan', 'ceilinglamp', 'mat', 'balanceball', 'mousepad', 'candle', 'wall', 'nightstand', 'newspaper', 'doorjamb', 'mouse', 'light', 'clothes_skirt']
Task:Browse internet
response:
{
"thoughts": Turn on computer, pull up browser, browse the internet,
"actions": ['[WALK] <home_office>', '[WALK] <computer>', '[FIND] <computer>', '[SWITCHON] <computer>', '[TURNTO] <computer>', '[LOOKAT] <computer>']
}
|
Scene_Objects:['pot', 'filing_cabinet', 'microwave', 'chair', 'freezer', 'drawing', 'knife', 'wallshelf', 'bathroom_counter', 'floor', 'ceiling', 'keyboard', 'toaster', 'tvstand', 'shoe-shine_kit', 'faucet', 'window', 'mousepad', 'couch', 'wall', 'powersocket', 'photoframe', 'coffe_maker', 'television', 'curtain', 'coffee_pot', 'towel_rack', 'sink', 'light', 'laser_pointer']
Task:Gaze out window
response:
| {
"thoughts": "Look at the yard from the window",
"actions": [
"[WALK] <home_office>",
"[WALK] <window>",
"[FIND] <window>",
"[TURNTO] <window>",
"[LOOKAT] <window>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['pot', 'tablelamp', 'chair', 'freezer', 'walllamp', 'milk', 'floor', 'keyboard', 'table_cloth', 'stovefan', 'mat', 'tvstand', 'faucet', 'window', 'couch', 'clothes_hat', 'pillow', 'dining_room', 'toy', 'basket_for_clothes', 'coffe_maker', 'video_game_controller', 'tray', 'doorjamb', 'kettle', 'dresser', 'cupboard', 'clothes_dress', 'sink', 'table']
Task:Pick up toys
response:
{
"thoughts": Bend down and pick up the toys off of the floor.,
"actions": ['[WALK] <bedroom>', '[WALK] <toy>', '[FIND] <toy>', '[GRAB] <toy>', '[WALK] <dresser>', '[OPEN] <dresser>', '[PUTIN] <toy> <dresser>', '[FIND] <toy>', '[GRAB] <toy>', '[WALK] <dresser>', '[PUTIN] <toy> <dresser>', '[FIND] <toy>', '[WALK] <dresser>', '[GRAB] <toy>', '[PUTIN] <toy> <dresser>', '[CLOSE] <dresser>']
}
|
Scene_Objects:['tea', 'microwave', 'toilet', 'freezer', 'walllamp', 'ground_coffee', 'blow_dryer', 'knifeblock', 'cpuscreen', 'bedroom', 'computer', 'stovefan', 'faucet', 'mousepad', 'food_peanut_butter', 'pillow', 'wall', 'oven', 'love_seat', 'newspaper', 'television', 'after_shave', 'doorjamb', 'mouse', 'kitchen_counter', 'dishrack', 'dresser', 'cupboard', 'conditioner', 'bathtub']
Task:Read
response:
| {
"thoughts": "Reading a magazine",
"actions": [
"[WALK] <home_office>",
"[WALK] <newspaper>",
"[FIND] <newspaper>",
"[GRAB] <newspaper>",
"[FIND] <love_seat>",
"[SIT] <love_seat>",
"[READ] <newspaper>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['food_ice_cream', 'measuring_cup', 'freezer', 'nail_polish', 'food_carrot', 'drawing', 'cpuscreen', 'kitchen_cabinet', 'floor', 'ceiling', 'desk', 'keyboard', 'computer', 'ceilinglamp', 'plate', 'mat', 'mousepad', 'food_peanut_butter', 'pillow', 'wall', 'powersocket', 'bench', 'food_vegetable', 'food_food', 'doorjamb', 'mouse', 'door', 'light', 'table', 'scrabble']
Task:Fix snack
response:
{
"thoughts": Open fridge. Take out carrots and other vegetables. Put vegetables on a plate. Put away carrots and celery. Close fridge.,
"actions": ['[WALK] <dining_room>', '[WALK] <kitchen_cabinet>', '[FIND] <kitchen_cabinet>', '[OPEN] <kitchen_cabinet>', '[FIND] <plate>', '[GRAB] <plate>', '[CLOSE] <kitchen_cabinet>', '[WALK] <freezer>', '[OPEN] <freezer>', '[FIND] <food_carrot>', '[GRAB] <food_carrot>', '[PUTBACK] <food_carrot> <plate>', '[FIND] <food_vegetable>', '[GRAB] <food_vegetable>', '[PUTBACK] <food_vegetable> <plate>', '[CLOSE] <freezer>', '[WALK] <table>', '[PUTBACK] <plate> <table>']
}
|
Scene_Objects:['pot', 'tablelamp', 'microwave', 'walllamp', 'detergent', 'knifeblock', 'floor', 'ceiling', 'keyboard', 'food_cake', 'computer', 'mat', 'faucet', 'mousepad', 'bowl', 'wall', 'nightstand', 'oven', 'bench', 'box', 'check', 'coffee_filter', 'tray', 'doorjamb', 'towel_rack', 'kitchen_counter', 'cupboard', 'sink', 'bathroom_cabinet', 'home_office']
Task:Wipe down sink
response:
| {
"thoughts": "Use a cloth to wipe the edge of the sink. Then wipe the bowl of the sink as well.",
"actions": [
"[WALK] <dining_room>",
"[WALK] <rag>",
"[FIND] <rag>",
"[GRAB] <rag>",
"[WALK] <sink>",
"[WIPE] <sink>",
"[PUTOBJBACK] <rag>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['pot', 'tea', 'microwave', 'bathroom', 'trashcan', 'orchid', 'knife', 'drawing', 'clothes_underwear', 'floor', 'ceiling', 'envelope', 'mat', 'mousepad', 'couch', 'pillow', 'wall', 'nightstand', 'novel', 'food_food', 'food_donut', 'curtain', 'doorjamb', 'kitchen_counter', 'hair', 'door', 'sink', 'table', 'light', 'dishwasher']
Task:Put out flowers
response:
{
"thoughts": Place a flower centrepiece on the entrance hall table.,
"actions": ['[WALK] <home_office>', '[WALK] <centerpiece>', '[FIND] <centerpiece>', '[GRAB] <centerpiece>', '[WALK] <table>', '[PUTBACK] <centerpiece> <table>']
}
|
Scene_Objects:['tablelamp', 'microwave', 'walllamp', 'food_carrot', 'orchid', 'drawing', 'wallshelf', 'floor', 'ceiling', 'phone', 'bookshelf', 'mat', 'tvstand', 'ceilinglamp', 'shower', 'window', 'couch', 'pillow', 'wall', 'powersocket', 'bench', 'photoframe', 'television', 'curtain', 'doorjamb', 'cards', 'cupboard', 'door', 'light', 'table']
Task:Watch TV
response:
| {
"thoughts": "Sit on couch and watch the tv",
"actions": [
"[WALK] <home_office>",
"[WALK] <television>",
"[FIND] <television>",
"[SWITCHON] <television>",
"[FIND] <couch>",
"[SIT] <couch>",
"[TURNTO] <television>",
"[WATCH] <television>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['pot', 'filing_cabinet', 'crayon', 'microwave', 'walllamp', 'dough', 'knifeblock', 'wallshelf', 'floor', 'phone', 'shampoo', 'desk', 'keyboard', 'bookmark', 'mat', 'shoe-shine_kit', 'floor_lamp', 'mousepad', 'pillow', 'powersocket', 'novel', 'curtain', 'centerpiece', 'kitchen_counter', 'door', 'bathtub', 'home_office', 'table', 'clothes_skirt', 'colander']
Task:Read book
response:
{
"thoughts": Turn a light on and read a book in the bedroom.,
"actions": ['[WALK] <bedroom>', '[WALK] <floor_lamp>', '[FIND] <floor_lamp>', '[SWITCHON] <floor_lamp>', '[FIND] <novel>', '[GRAB] <novel>', '[READ] <novel>']
}
|
Scene_Objects:['tablelamp', 'bathroom', 'drawing', 'knifeblock', 'floor', 'character', 'keyboard', 'stovefan', 'mat', 'ceilinglamp', 'stereo', 'shower', 'window', 'toaster', 'wall', 'nightstand', 'box', 'food_cheese', 'bed', 'after_shave', 'lighter', 'diary', 'doorjamb', 'coffee_pot', 'kitchen_counter', 'cupboard', 'food_steak', 'light', 'dishwasher', 'clothes_shirt']
Task:Listen to music
response:
| {
"thoughts": "Turn on radio, put on favorite channel, listen to music.",
"actions": [
"[WALK] <bedroom>",
"[WALK] <stereo>",
"[FIND] <stereo>",
"[SWITCHON] <stereo>",
"[FIND] <bed>",
"[LIE] <bed>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['glue', 'pot', 'tablelamp', 'detergent', 'cat', 'knife', 'drawing', 'hanger', 'bathroom_counter', 'desk', 'food_cake', 'purse', 'envelope', 'piano_bench', 'stovefan', 'toaster', 'stereo', 'balanceball', 'floor_lamp', 'powersocket', 'love_seat', 'newspaper', 'bed', 'photoframe', 'video_game_controller', 'diary', 'maindoor', 'light', 'soap', 'board_game']
Task:Read
response:
{
"thoughts": Sit on an easy chair. Turn on music. Take a book and read. Turn on lamp at night.,
"actions": ['[WALK] <home_office>', '[WALK] <stereo>', '[FIND] <stereo>', '[SWITCHON] <stereo>', '[FIND] <newspaper>', '[GRAB] <newspaper>', '[FIND] <love_seat>', '[SIT] <love_seat>', '[FIND] <floor_lamp>', '[SWITCHON] <floor_lamp>', '[READ] <newspaper>']
}
|
Scene_Objects:['filing_cabinet', 'tablelamp', 'knifeblock', 'needle', 'wallshelf', 'bathroom_counter', 'phone', 'desk', 'food_cake', 'computer', 'ceilinglamp', 'tvstand', 'faucet', 'mat', 'window', 'shower', 'stereo', 'food_peanut_butter', 'clothes_pants', 'pillow', 'bench', 'newspaper', 'bed', 'coffe_maker', 'television', 'lighter', 'dresser', 'cupboard', 'clothes_dress', 'table']
Task:Get ready for school
response:
| {
"thoughts": "Take clothes out of dressers and get dressed for school.",
"actions": [
"[WALK] <bedroom>",
"[WALK] <dresser>",
"[OPEN] <dresser>",
"[FIND] <clothes_dress>",
"[GRAB] <clothes_dress>",
"[CLOSE] <dresser>",
"[PUTON] <clothes_dress>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['microwave', 'chair', 'jelly', 'walllamp', 'orchid', 'wallshelf', 'cpuscreen', 'bathroom_counter', 'floor', 'ceiling', 'desk', 'keyboard', 'bookshelf', 'envelope', 'computer', 'stovefan', 'ceilinglamp', 'tvstand', 'faucet', 'picture', 'mousepad', 'wall', 'powersocket', 'television', 'tray', 'doorjamb', 'towel_rack', 'mouse', 'bathroom_cabinet', 'table']
Task:Work
response:
{
"thoughts": Go into the office and sit down at the desk in front of the computer. Turn on the computer and look at it while clicking with the mouse and typing with the keyboard.,
"actions": ['[WALK] <home_office>', '[WALK] <chair>', '[FIND] <chair>', '[SIT] <chair>', '[FIND] <computer>', '[SWITCHON] <computer>', '[TURNTO] <computer>', '[LOOKAT] <computer>', '[FIND] <keyboard>', '[TYPE] <keyboard>']
}
|
Scene_Objects:['walllamp', 'drawing', 'orchid', 'wallshelf', 'floor', 'ceiling', 'bookshelf', 'remote_control', 'mat', 'tvstand', 'ceilinglamp', 'shower', 'window', 'couch', 'pillow', 'wall', 'powersocket', 'photoframe', 'newspaper', 'television', 'curtain', 'doorjamb', 'towel_rack', 'kitchen_counter', 'dresser', 'door', 'napkin', 'spoon', 'table', 'light']
Task:Watch TV
response:
| {
"thoughts": "Sit on the couch and use the remote control to change channels",
"actions": [
"[WALK] <home_office>",
"[WALK] <remote_control>",
"[FIND] <remote_control>",
"[GRAB] <remote_control>",
"[FIND] <couch>",
"[SIT] <couch>",
"[FIND] <television>",
"[TURNTO] <television>",
"[POINTAT] <television>",
"[SWITCHON] <television>",
"[PUTBACK] <remote_control> <couch>",
"[WATCH] <television>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['glue', 'pot', 'toilet', 'chair', 'freezer', 'band-aids', 'bathroom', 'food_carrot', 'drawing', 'floor', 'mat', 'window', 'couch', 'clothes_hat', 'form', 'pillow', 'bench', 'food_food', 'check', 'coffe_maker', 'toilet_paper', 'ice', 'centerpiece', 'doorjamb', 'cupboard', 'home_office', 'table', 'light', 'bathroom_cabinet', 'scrabble']
Task:Eat dinner
response:
{
"thoughts": In our dining room, my family eats dinner. Once the food has been prepared, we gather at the dining room table and prepare to eat.,
"actions": ['[WALK] <dining_room>', '[WALK] <food_food>', '[FIND] <food_food>', '[GRAB] <food_food>', '[WALK] <dining_room>', '[WALK] <table>', '[PUTBACK] <food_food> <table>', '[WALK] <dining_room>', '[WALK] <food_food>', '[FIND] <food_food>', '[GRAB] <food_food>', '[WALK] <dining_room>', '[WALK] <table>', '[PUTBACK] <food_food> <table>', '[WALK] <chair>', '[PULL] <chair>', '[SIT] <chair>', '[TURNTO] <food_food>', '[LOOKAT] <food_food>', '[TURNTO] <food_food>', '[LOOKAT] <food_food>']
}
|
Scene_Objects:['pot', 'food_bread', 'chair', 'food_carrot', 'cpuscreen', 'bathroom_counter', 'character', 'bedroom', 'keyboard', 'food_bacon', 'bookmark', 'stovefan', 'ceilinglamp', 'tvstand', 'picture', 'window', 'pillow', 'bench', 'love_seat', 'photoframe', 'novel', 'coin', 'curtain', 'towel_rack', 'mouse', 'pencil', 'sink', 'light', 'table', 'instrument_violin']
Task:Read book
response:
| {
"thoughts": "Obtain a book from the book shelf. Sit in the easy chair. Read the book. Mark the stopping point with a bookmark. Put the book back on the book shelf.",
"actions": [
"[WALK] <home_office>",
"[WALK] <novel>",
"[FIND] <novel>",
"[GRAB] <novel>",
"[FIND] <love_seat>",
"[SIT] <love_seat>",
"[READ] <novel>",
"[FIND] <bookmark>",
"[GRAB] <bookmark>",
"[PUTBACK] <bookmark> <novel>",
"[PUTOBJBACK] <novel>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['tablelamp', 'microwave', 'toilet', 'freezer', 'walllamp', 'trashcan', 'drawing', 'cpuscreen', 'wallshelf', 'hanger', 'floor', 'ceiling', 'character', 'food_bacon', 'mat', 'ceilinglamp', 'shower', 'wall', 'wall_clock', 'cup', 'photoframe', 'bed', 'food_food', 'tray', 'doorjamb', 'mouse', 'kitchen_counter', 'cupboard', 'door', 'coffee']
Task:Drink
response:
{
"thoughts": Walk into kitchen, look for cup, walk to cup, look for refrigerator, walk to refrigerator, open door, grab water, pour water into cup,
"actions": ['[WALK] <dining_room>', '[WALK] <cup>', '[FIND] <cup>', '[TURNTO] <cup>', '[LOOKAT] <cup>', '[WALK] <cup>', '[FIND] <freezer>', '[TURNTO] <freezer>', '[LOOKAT] <freezer>', '[WALK] <freezer>', '[GRAB] <cup>', '[WALK] <water>', '[FIND] <water>', '[GRAB] <water>', '[POUR] <water> <cup>']
}
|
Scene_Objects:['food_pizza', 'chair', 'freezer', 'walllamp', 'food_carrot', 'knifeblock', 'floor', 'ceiling', 'desk', 'keyboard', 'computer', 'toaster', 'mat', 'faucet', 'window', 'oil', 'couch', 'wall', 'bench', 'powersocket', 'photoframe', 'bed', 'television', 'curtain', 'after_shave', 'towel_rack', 'kitchen_counter', 'bathtub', 'sink', 'light']
Task:Open bathroom window
response:
| {
"thoughts": "Make sure bathroom window is unlocked. Lift up bathroom window so it is open.",
"actions": [
"[WALK] <bathroom>",
"[WALK] <window>",
"[FIND] <window>",
"[OPEN] <window>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['dry_pasta', 'tablelamp', 'chair', 'walllamp', 'orchid', 'floor', 'ceiling', 'phone', 'bookshelf', 'remote_control', 'computer', 'ceilinglamp', 'tvstand', 'window', 'oil', 'closetdrawer', 'couch', 'pillow', 'dining_room', 'nightstand', 'wall', 'oven', 'bed', 'coffe_maker', 'television', 'curtain', 'dresser', 'door', 'instrument_guitar', 'table']
Task:Watch TV
response:
{
"thoughts": I go to the living room, switch on the tv, take the remote and sit on the sofa.,
"actions": ['[WALK] <home_office>', '[WALK] <remote_control>', '[FIND] <remote_control>', '[GRAB] <remote_control>', '[FIND] <couch>', '[SIT] <couch>', '[FIND] <television>', '[SWITCHON] <television>', '[TURNTO] <television>', '[WATCH] <television>']
}
|
Scene_Objects:['walllamp', 'drawing', 'orchid', 'wallshelf', 'knifeblock', 'knife', 'bathroom_counter', 'floor', 'ceiling', 'bookshelf', 'mat', 'tvstand', 'ceilinglamp', 'shower', 'window', 'couch', 'pillow', 'wall', 'powersocket', 'photoframe', 'light_bulb', 'television', 'curtain', 'doorjamb', 'cd', 'pajamas', 'door', 'table', 'light', 'dishwasher']
Task:Watch TV
response:
| {
"thoughts": "Sitting on the couch watching my favorite television show.",
"actions": [
"[WALK] <home_office>",
"[WALK] <couch>",
"[FIND] <couch>",
"[SIT] <couch>",
"[FIND] <television>",
"[TURNTO] <television>",
"[WATCH] <television>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['sheets', 'tablelamp', 'microwave', 'walllamp', 'milk', 'clothes_underwear', 'character', 'bedroom', 'keyboard', 'clothes_socks', 'candle', 'clothes_pants', 'wall', 'nightstand', 'powersocket', 'basket_for_clothes', 'photoframe', 'bed', 'food_food', 'curtain', 'diary', 'shoes', 'headset', 'laundry_detergent', 'sink', 'home_office', 'table', 'colander', 'clothes_shirt', 'washing_machine']
Task:Wash clothes
response:
{
"thoughts": I grab the basket for clothes from the bedroom. I carry it and walk to the washing machine. I open the washing machine. I grab the dirty clothes. I put the clothes in the washing machine. I grab the laundry detergent and put it in the washing machine. I close the washing machine. I switch on the washing machine.,
"actions": ['[WALK] <bedroom>', '[WALK] <basket_for_clothes>', '[FIND] <basket_for_clothes>', '[GRAB] <basket_for_clothes>', '[WALK] <washing_machine>', '[DROP] <basket_for_clothes>', '[OPEN] <washing_machine>', '[FIND] <clothes_pants>', '[GRAB] <clothes_pants>', '[PUTBACK] <clothes_pants> <washing_machine>', '[FIND] <clothes_shirt>', '[GRAB] <clothes_shirt>', '[PUTBACK] <clothes_shirt> <washing_machine>', '[FIND] <clothes_underwear>', '[GRAB] <clothes_underwear>', '[PUTBACK] <clothes_underwear> <washing_machine>', '[FIND] <clothes_socks>', '[GRAB] <clothes_socks>', '[PUTBACK] <clothes_socks> <washing_machine>', '[FIND] <laundry_detergent>', '[GRAB] <laundry_detergent>', '[PUTBACK] <laundry_detergent> <washing_machine>', '[GRAB] <laundry_detergent>', '[PUTOBJBACK] <laundry_detergent>', '[CLOSE] <washing_machine>', '[SWITCHON] <washing_machine>']
}
|
Scene_Objects:['pot', 'tea', 'microwave', 'chair', 'freezer', 'walllamp', 'milk', 'knifeblock', 'knife', 'drawing', 'floor', 'ceiling', 'shampoo', 'desk', 'bookshelf', 'stovefan', 'toaster', 'faucet', 'food_jam', 'pillow', 'wall', 'oven', 'powersocket', 'cup', 'photoframe', 'tray', 'kitchen_counter', 'cupboard', 'controller', 'sink']
Task:Get something to drink
response:
| {
"thoughts": "Walk into the kitchen and to the cupboard to get a cup, walk to the fridge to look for a beverage to enjoy, open the freezer and get ice, pour beverage and enjoy",
"actions": [
"[WALK] <dining_room>",
"[WALK] <cupboard>",
"[FIND] <cupboard>",
"[OPEN] <cupboard>",
"[FIND] <cup>",
"[GRAB] <cup>",
"[CLOSE] <cupboard>",
"[WALK] <freezer>",
"[OPEN] <freezer>",
"[FIND] <milk>",
"[GRAB] <milk>",
"[CLOSE] <freezer>",
"[POUR] <milk> <cup>",
"[DRINK] <cup>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['pot', 'tea', 'chair', 'detergent', 'toothbrush', 'orchid', 'knifeblock', 'cpuscreen', 'kitchen_cabinet', 'floor', 'phone', 'bedroom', 'keyboard', 'ceilinglamp', 'faucet', 'window', 'dish_soap', 'food_jam', 'food_orange', 'pillow', 'wall', 'oven', 'bed', 'novel', 'coffe_maker', 'kitchen_counter', 'dishrack', 'cupboard', 'table', 'dishwasher']
Task:Wash dishes with dishwasher
response:
{
"thoughts": I walk to the kitchen. I open the dishwasher. I open the kitchen cabinet. I reach inside the cabinet and grab the box of dish soap. I add some soap to the open dishwasher. I put the box of dish soap back in the kitchen cabinet. I close the kitchen cabinet and then I close the dishwasher. I turn the dishwasher on and wait.,
"actions": ['[WALK] <dining_room>', '[WALK] <dishwasher>', '[FIND] <dishwasher>', '[OPEN] <dishwasher>', '[FIND] <kitchen_cabinet>', '[OPEN] <kitchen_cabinet>', '[FIND] <dish_soap>', '[GRAB] <dish_soap>', '[PUTBACK] <dish_soap> <dishwasher>', '[GRAB] <dish_soap>', '[PUTIN] <dish_soap> <kitchen_cabinet>', '[CLOSE] <kitchen_cabinet>', '[CLOSE] <dishwasher>', '[SWITCHON] <dishwasher>']
}
|
Scene_Objects:['pot', 'filing_cabinet', 'dog', 'bathroom', 'knife', 'phone', 'character', 'desk', 'clothes_jacket', 'mat', 'ceilinglamp', 'window', 'mousepad', 'pillow', 'dining_room', 'photoframe', 'keys', 'coffe_maker', 'food_food', 'check', 'ice', 'tray', 'towel_rack', 'mouse', 'shoes', 'cupboard', 'bathtub', 'home_office', 'light', 'bathroom_cabinet']
Task:Get ready to leave
response:
| {
"thoughts": "Get your jacket and shoes on, locate the keys off the side table, walk out the front door and secure building with keys when you leave",
"actions": [
"[WALK] <home_office>",
"[WALK] <shoes>",
"[FIND] <shoes>",
"[GRAB] <shoes>",
"[PUTON] <shoes>",
"[FIND] <keys>",
"[GRAB] <keys>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['sheets', 'pot', 'mop', 'dog', 'jelly', 'drawing', 'knife', 'bathroom_counter', 'floor', 'ceiling', 'longboard', 'character', 'computer', 'stovefan', 'faucet', 'oil', 'dining_room', 'wall', 'nightstand', 'oven', 'cutting_board', 'photoframe', 'check', 'tray', 'sink', 'napkin', 'bathroom_cabinet', 'soap', 'dirt', 'hands_both']
Task:Wash hands
response:
{
"thoughts": go to the bathroom and wash your hands,
"actions": ['[WALK] <bathroom>', '[WALK] <sink>', '[FIND] <faucet>', '[SWITCHON] <faucet>', '[FIND] <soap>', '[SQUEEZE] <soap>', '[FIND] <hands_both>', '[WASH] <hands_both>']
}
|
Scene_Objects:['sheets', 'toilet', 'trashcan', 'drawing', 'orchid', 'knifeblock', 'floor', 'ceiling', 'bedroom', 'keyboard', 'stovefan', 'mousepad', 'couch', 'pillow', 'wall', 'wall_clock', 'powersocket', 'newspaper', 'coffee_table', 'novel', 'bed', 'oven_mitts', 'doorjamb', 'pencil', 'chessboard', 'dresser', 'door', 'food_steak', 'table', 'light']
Task:Read magazine
response:
| {
"thoughts": "Get magazine off coffee table. Sit in arm chair. Open magazine to first page. Read until finished.",
"actions": [
"[WALK] <home_office>",
"[WALK] <coffee_table>",
"[FIND] <newspaper>",
"[GRAB] <newspaper>",
"[FIND] <couch>",
"[SIT] <couch>",
"[READ] <newspaper>",
"[PUTBACK] <newspaper> <coffee_table>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['spectacles', 'filing_cabinet', 'chair', 'walllamp', 'bathroom', 'cpuscreen', 'floor', 'ceiling', 'phone', 'desk', 'keyboard', 'bookshelf', 'computer', 'ceilinglamp', 'tvstand', 'mousepad', 'dining_room', 'wall', 'powersocket', 'oven', 'newspaper', 'bed', 'television', 'doorjamb', 'mouse', 'kitchen_counter', 'shoes', 'cupboard', 'bathroom_cabinet', 'table']
Task:Use computer
response:
{
"thoughts": Sit at desk, use the computer mouse,
"actions": ['[WALK] <home_office>', '[WALK] <chair>', '[FIND] <chair>', '[SIT] <chair>', '[FIND] <computer>', '[SWITCHON] <computer>', '[FIND] <mouse>', '[GRAB] <mouse>', '[FIND] <mousepad>', '[PUTBACK] <mouse> <mousepad>', '[TURNTO] <computer>', '[POINTAT] <computer>']
}
|
Scene_Objects:['tablelamp', 'blender', 'chair', 'bathroom', 'detergent', 'drawing', 'cpuscreen', 'floor', 'phone', 'desk', 'mat', 'tvstand', 'clothes_pants', 'wall', 'nightstand', 'powersocket', 'photoframe', 'bed', 'novel', 'check', 'keys', 'tray', 'after_shave', 'curtain', 'doorjamb', 'kitchen_counter', 'cupboard', 'door', 'bathtub', 'tooth_paste']
Task:Read
response:
| {
"thoughts": "I like to read books in my bedroom.",
"actions": [
"[WALK] <bedroom>",
"[WALK] <desk>",
"[FIND] <novel>",
"[GRAB] <novel>",
"[WALK] <bed>",
"[LIE] <bed>",
"[READ] <novel>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['spectacles', 'tablelamp', 'microwave', 'chair', 'wallshelf', 'hanger', 'floor', 'phone', 'food_bacon', 'piano_bench', 'stovefan', 'ceilinglamp', 'shower', 'mousepad', 'closetdrawer', 'powersocket', 'television', 'coin', 'food_donut', 'ice', 'rag', 'pencil', 'shoes', 'dresser', 'clothes_dress', 'clothes_scarf', 'sink', 'bathroom_cabinet', 'light', 'home_office']
Task:Hang up jacket
response:
{
"thoughts": I put clothes on hangers and hang them in the closet,
"actions": ['[WALK] <bedroom>', '[WALK] <clothes_dress>', '[FIND] <clothes_dress>', '[GRAB] <clothes_dress>', '[WALK] <dresser>', '[FIND] <hanger>', '[OPEN] <dresser>', '[PUTIN] <clothes_dress> <dresser>', '[FIND] <clothes_dress>', '[GRAB] <clothes_dress>', '[WALK] <dresser>', '[PUTIN] <clothes_dress> <dresser>', '[CLOSE] <dresser>']
}
|
Scene_Objects:['crayon', 'chair', 'freezer', 'blow_dryer', 'bathroom', 'cat', 'trashcan', 'orchid', 'bathroom_counter', 'floor', 'desk', 'keyboard', 'shower', 'window', 'mousepad', 'microphone', 'dining_room', 'wall', 'pasta', 'bench', 'oven', 'newspaper', 'check', 'television', 'tray', 'curtain', 'towel_rack', 'mouse', 'home_office', 'bathroom_cabinet']
Task:Read newspaper
response:
| {
"thoughts": "Pull out dining room chair. Sit in dining room chair. Open newspaper to section you'd like to read. Read section. Keep reading all sections you'd like.",
"actions": [
"[WALK] <dining_room>",
"[WALK] <newspaper>",
"[FIND] <newspaper>",
"[GRAB] <newspaper>",
"[FIND] <chair>",
"[SIT] <chair>",
"[READ] <newspaper>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['tablelamp', 'food_ice_cream', 'measuring_cup', 'freezer', 'ceilingfan', 'cpuscreen', 'longboard', 'ceiling', 'character', 'bedroom', 'toaster', 'stereo', 'balanceball', 'oil', 'couch', 'food_orange', 'pillow', 'food_kiwi', 'nightstand', 'photoframe', 'coffe_maker', 'rag', 'doorjamb', 'towel_rack', 'shoes', 'controller', 'napkin', 'bathroom_cabinet', 'laser_pointer', 'dirt']
Task:Pick up phone
response:
{
"thoughts": Go to the living room. Find the telephone. Lift up the receiver.,
"actions": ['[WALK] <home_office>', '[WALK] <phone>', '[FIND] <phone>', '[GRAB] <phone>']
}
|
Scene_Objects:['walllamp', 'knifeblock', 'knife', 'cpuscreen', 'ceiling', 'bookshelf', 'purse', 'face', 'mail', 'mat', 'deck_of_cards', 'tvstand', 'homework', 'pillow', 'dining_room', 'nightstand', 'powersocket', 'box', 'photoframe', 'bed', 'novel', 'check', 'coffe_maker', 'mirror', 'tray', 'hair', 'clothes_dress', 'sink', 'bathroom_cabinet', 'home_office']
Task:Check appearance in mirror
response:
| {
"thoughts": "Walk to full-length mirror. Look in mirror. Make sure everything about your appearance is adequate. Adjust clothes as needed.",
"actions": [
"[WALK] <bedroom>",
"[WALK] <mirror>",
"[FIND] <mirror>",
"[TURNTO] <mirror>",
"[LOOKAT] <mirror>",
"[FIND] <face>",
"[TURNTO] <face>",
"[LOOKAT] <face>",
"[FIND] <hair>",
"[TURNTO] <hair>",
"[LOOKAT] <hair>",
"[FIND] <clothes_dress>",
"[TURNTO] <clothes_dress>",
"[LOOKAT] <clothes_dress>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['food_pizza', 'dog', 'bathroom', 'trashcan', 'drawing', 'orchid', 'wallshelf', 'floor', 'ceiling', 'character', 'desk', 'shower', 'electric_shaver', 'window', 'couch', 'pillow', 'wall', 'nightstand', 'newspaper', 'bed', 'food_food', 'doorjamb', 'towel_rack', 'mouse', 'kitchen_counter', 'cupboard', 'door', 'bathroom_cabinet', 'table', 'light']
Task:Relax on sofa
response:
{
"thoughts": I walk to the living room and find the sofa. I sit on the sofa and get comfortable.,
"actions": ['[WALK] <home_office>', '[WALK] <couch>', '[FIND] <couch>', '[SIT] <couch>']
}
|
Scene_Objects:['tablelamp', 'chair', 'knifeblock', 'drawing', 'cpuscreen', 'orchid', 'floor', 'phone', 'bedroom', 'keyboard', 'mat', 'tvstand', 'floor_lamp', 'teeth', 'couch', 'pillow', 'wall', 'nightstand', 'powersocket', 'bed', 'ice', 'tray', 'after_shave', 'doorjamb', 'centerpiece', 'mouse', 'dresser', 'light', 'coffee', 'creditcard']
Task:Turn on lights
response:
| {
"thoughts": "Walk to your lamp, switch it on, and lay on your bed",
"actions": [
"[WALK] <bedroom>",
"[WALK] <floor_lamp>",
"[FIND] <floor_lamp>",
"[TURNTO] <floor_lamp>",
"[SWITCHON] <floor_lamp>",
"[FIND] <bed>",
"[LIE] <bed>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['filing_cabinet', 'juice', 'freezer', 'blow_dryer', 'trashcan', 'drawing', 'cpuscreen', 'longboard', 'floor', 'bookmark', 'ceilinglamp', 'mat', 'window', 'couch', 'microphone', 'clothes_hat', 'pillow', 'wall', 'bench', 'cup', 'keys', 'television', 'towel_rack', 'cd', 'cupboard', 'conditioner', 'door', 'home_office', 'table', 'vase']
Task:Relax on sofa
response:
{
"thoughts": Go to the living room and relax on the sofa with a book,
"actions": ['[RUN] <home_office>', '[WALK] <novel>', '[FIND] <novel>', '[GRAB] <novel>', '[WALK] <couch>', '[SIT] <couch>', '[TURNTO] <novel>', '[LOOKAT] <novel>', '[FIND] <bookmark>', '[GRAB] <bookmark>', '[DROP] <bookmark>', '[TURNTO] <novel>', '[LOOKAT] <novel>', '[READ] <novel>', '[READ] <novel>', '[READ] <novel>']
}
|
Scene_Objects:['pot', 'tablelamp', 'microwave', 'chair', 'freezer', 'cat', 'cpuscreen', 'bathroom_counter', 'floor', 'phone', 'character', 'stovefan', 'mat', 'tvstand', 'stereo', 'iron', 'food_kiwi', 'powersocket', 'newspaper', 'novel', 'coffe_maker', 'tray', 'doorjamb', 'towel_rack', 'mouse', 'kitchen_counter', 'conditioner', 'bathtub', 'home_office', 'electrical_outlet']
Task:Turn on radio
response:
| {
"thoughts": "Go to the radio. Switch on the power button. Tune the station. Set the volume. Listen to the music.",
"actions": [
"[WALK] <home_office>",
"[WALK] <stereo>",
"[FIND] <stereo>",
"[FIND] <electrical_outlet>",
"[TURNTO] <electrical_outlet>",
"[PLUGIN] <stereo>",
"[SWITCHON] <electrical_outlet>",
"[SWITCHON] <stereo>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['blender', 'chair', 'shampoo', 'ceiling', 'folder', 'face', 'computer', 'ceilinglamp', 'tvstand', 'balanceball', 'picture', 'oil', 'food_orange', 'dining_room', 'bench', 'box', 'newspaper', 'check', 'video_game_controller', 'mirror', 'curtain', 'towel_rack', 'hair', 'clothes_dress', 'laundry_detergent', 'bathroom_cabinet', 'table', 'clothes_skirt', 'vase', 'scrabble']
Task:Check appearance in mirror
response:
{
"thoughts": Walk to full-length mirror. Look in mirror. Make sure everything about your appearance is adequate. Adjust clothes as needed.,
"actions": ['[WALK] <bedroom>', '[WALK] <mirror>', '[FIND] <mirror>', '[TURNTO] <mirror>', '[LOOKAT] <mirror>', '[FIND] <face>', '[TURNTO] <face>', '[LOOKAT] <face>', '[FIND] <hair>', '[TURNTO] <hair>', '[LOOKAT] <hair>', '[FIND] <clothes_dress>', '[TURNTO] <clothes_dress>', '[LOOKAT] <clothes_dress>']
}
|
Scene_Objects:['sheets', 'pot', 'filing_cabinet', 'tablelamp', 'mop', 'microwave', 'toilet', 'band-aids', 'bathroom', 'trashcan', 'wallshelf', 'character', 'computer', 'mat', 'faucet', 'couch', 'dining_room', 'wall', 'newspaper', 'oven_mitts', 'tray', 'curtain', 'rag', 'doorjamb', 'towel_rack', 'mouse', 'sink', 'bathroom_cabinet', 'table', 'scrabble']
Task:Shut front door
response:
| {
"thoughts": "Close the storm door. Lock the storm door.",
"actions": [
"[WALK] <home_office>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['tablelamp', 'freezer', 'walllamp', 'ground_coffee', 'trashcan', 'drawing', 'cpuscreen', 'floor', 'phone', 'ceiling', 'character', 'bookshelf', 'food_cake', 'mat', 'tvstand', 'shoe-shine_kit', 'homework', 'oil', 'couch', 'food_peanut_butter', 'pillow', 'dining_room', 'wall', 'bench', 'oven', 'powersocket', 'toilet_paper', 'kitchen_counter', 'door', 'tooth_paste']
Task:Pick up phone
response:
{
"thoughts": walk over and pick up the phone,
"actions": ['[WALK] <phone>', '[GRAB] <phone>']
}
|
Scene_Objects:['pot', 'tablelamp', 'microwave', 'walllamp', 'bathroom', 'knifeblock', 'drawing', 'wallshelf', 'floor', 'desk', 'keyboard', 'mat', 'tvstand', 'shower', 'faucet', 'stereo', 'wall', 'nightstand', 'powersocket', 'photoframe', 'television', 'toothbrush_holder', 'curtain', 'doorjamb', 'towel_rack', 'kitchen_counter', 'cupboard', 'bathtub', 'table', 'dishwasher']
Task:Put in chair
response:
| {
"thoughts": "Take a comfortable chair. Place it in the entrance hall",
"actions": [
"[WALK] <home_office>",
"[WALK] <chair>",
"[FIND] <chair>",
"[GRAB] <chair>",
"[WALK] <home_office>",
"[WALK] <floor>",
"[FIND] <floor>",
"[PUTBACK] <chair> <floor>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['microwave', 'toilet', 'freezer', 'walllamp', 'bathroom', 'orchid', 'drawing', 'wallshelf', 'hanger', 'bathroom_counter', 'ceiling', 'desk', 'computer', 'ceilinglamp', 'shower', 'window', 'mousepad', 'closetdrawer', 'pillow', 'wall', 'wall_clock', 'powersocket', 'photoframe', 'bed', 'food_food', 'mouse', 'vacuum_cleaner', 'bathtub', 'sink', 'table']
Task:Pick up phone
response:
{
"thoughts": Walk into the living room, walk over to phone, grab the phone, put it to your ear .,
"actions": ['[WALK] <home_office>', '[WALK] <phone>', '[GRAB] <phone>']
}
|
Scene_Objects:['walllamp', 'orchid', 'wallshelf', 'floor', 'ceiling', 'bedroom', 'bookshelf', 'mat', 'tvstand', 'ceilinglamp', 'shower', 'window', 'mousepad', 'dining_room', 'wall', 'nightstand', 'powersocket', 'photoframe', 'keys', 'television', 'curtain', 'food_onion', 'doorjamb', 'mouse', 'dresser', 'pajamas', 'sink', 'home_office', 'light', 'instrument_guitar']
Task:Turn off TV
response:
| {
"thoughts": "Walk over to tv and press power button",
"actions": [
"[WALK] <home_office>",
"[WALK] <television>",
"[SWITCHOFF] <television>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['sheets', 'tablelamp', 'blender', 'freezer', 'walllamp', 'dough', 'orchid', 'drawing', 'floor', 'ceiling', 'remote_control', 'ceilinglamp', 'tvstand', 'mat', 'shower', 'window', 'picture', 'couch', 'pillow', 'bowl', 'wall', 'powersocket', 'television', 'curtain', 'doorjamb', 'coffee_pot', 'dishrack', 'door', 'sink', 'table']
Task:Watch TV
response:
{
"thoughts": Turn on the TV, sit on the couch, find a show, and watch it.,
"actions": ['[WALK] <home_office>', '[WALK] <television>', '[FIND] <television>', '[FIND] <remote_control>', '[FIND] <couch>', '[SIT] <couch>', '[SWITCHON] <television>', '[TURNTO] <television>', '[LOOKAT] <television>']
}
|
Scene_Objects:['spectacles', 'tablelamp', 'blender', 'microwave', 'chair', 'walllamp', 'cpuscreen', 'bathroom_counter', 'ceiling', 'character', 'mail', 'food_bacon', 'mat', 'stereo', 'faucet', 'homework', 'shower', 'dining_room', 'wall', 'oven', 'keys', 'television', 'food_onion', 'coffee_pot', 'towel_rack', 'kitchen_counter', 'laundry_detergent', 'bag', 'spoon', 'table']
Task:Put away keys
response:
| {
"thoughts": "Return keys into a bag",
"actions": [
"[WALK] <home_office>",
"[WALK] <keys>",
"[FIND] <keys>",
"[GRAB] <keys>",
"[FIND] <bag>",
"[OPEN] <bag>",
"[PUTBACK] <keys> <bag>",
"[CLOSE] <bag>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['tablelamp', 'drawing', 'wallshelf', 'floor', 'ceiling', 'shampoo', 'character', 'desk', 'bookshelf', 'mat', 'deck_of_cards', 'couch', 'pillow', 'wall', 'powersocket', 'cup', 'food_cheese', 'photoframe', 'bed', 'coffee_filter', 'wine', 'curtain', 'doorjamb', 'kitchen_counter', 'pajamas', 'door', 'food_steak', 'bag', 'light', 'electrical_outlet']
Task:Turn on light
response:
{
"thoughts": I walk to the living room. I find the electric outlet. I switch on the light.,
"actions": ['[WALK] <home_office>', '[WALK] <electrical_outlet>', '[FIND] <electrical_outlet>', '[FIND] <light>', '[SWITCHON] <light>']
}
|
Scene_Objects:['spectacles', 'glue', 'tea', 'chair', 'freezer', 'walllamp', 'knifeblock', 'bathroom_counter', 'floor', 'phone', 'bedroom', 'keyboard', 'bookshelf', 'food_bacon', 'tvstand', 'faucet', 'window', 'mousepad', 'wall', 'powersocket', 'photoframe', 'coffe_maker', 'toilet_paper', 'diary', 'mouse', 'chessboard', 'spoon', 'home_office', 'table', 'light']
Task:Sit in chair
response:
| {
"thoughts": "I sit in a chair to rest and eat.",
"actions": [
"[WALK] <dining_room>",
"[WALK] <chair>",
"[FIND] <chair>",
"[PULL] <chair>",
"[SIT] <chair>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['glue', 'tablelamp', 'microwave', 'freezer', 'drawing', 'phone', 'bookshelf', 'toaster', 'stovefan', 'tvstand', 'mousepad', 'couch', 'clothes_hat', 'food_orange', 'wall', 'photoframe', 'clothes_gloves', 'bed', 'check', 'television', 'food_rice', 'tray', 'doorjamb', 'dresser', 'hairdryer', 'sink', 'bathtub', 'bathroom_cabinet', 'light', 'door']
Task:Listen to music
response:
{
"thoughts": I go to my bedroom, and then I switch on the radio,
"actions": ['[WALK] <bedroom>', '[WALK] <stereo>', '[FIND] <stereo>', '[SWITCHON] <stereo>']
}
|
Scene_Objects:['walllamp', 'nail_polish', 'trashcan', 'milk', 'drawing', 'wallshelf', 'bathroom_counter', 'ceiling', 'phone', 'bedroom', 'bookshelf', 'computer', 'faucet', 'pillow', 'wall', 'photoframe', 'toilet_paper', 'check', 'coffe_maker', 'wine', 'after_shave', 'towel_rack', 'comb', 'controller', 'sink', 'bathroom_cabinet', 'light', 'table', 'instrument_violin', 'teeth']
Task:Get toilet paper
response:
| {
"thoughts": "There are extra toilet papers in the bottom cabinet.",
"actions": [
"[WALK] <bathroom>",
"[WALK] <bathroom_cabinet>",
"[FIND] <bathroom_cabinet>",
"[OPEN] <bathroom_cabinet>",
"[FIND] <toilet_paper>",
"[GRAB] <toilet_paper>",
"[CLOSE] <bathroom_cabinet>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['pot', 'microwave', 'chair', 'freezer', 'wallshelf', 'cpuscreen', 'floor', 'phone', 'desk', 'remote_control', 'computer', 'button', 'toaster', 'mat', 'shower', 'food_peanut_butter', 'food_kiwi', 'nightstand', 'oven', 'photoframe', 'television', 'oven_mitts', 'tray', 'kettle', 'cards', 'pencil', 'kitchen_counter', 'towel_rack', 'bathtub', 'table']
Task:Change TV channel
response:
{
"thoughts": Walk into living room. Find television remote control. Pick up remote control. Press button.,
"actions": ['[WALK] <home_office>', '[WALK] <remote_control>', '[FIND] <remote_control>', '[GRAB] <remote_control>', '[FIND] <button>', '[PUSH] <button>']
}
|
Scene_Objects:['spectacles', 'pot', 'scissors', 'microwave', 'chair', 'freezer', 'drawing', 'orchid', 'cpuscreen', 'floor', 'ceiling', 'toaster', 'tvstand', 'faucet', 'dining_room', 'wall', 'bed', 'novel', 'food_food', 'mouthwash', 'rag', 'doorjamb', 'mouse', 'kitchen_counter', 'dresser', 'cupboard', 'home_office', 'bathroom_cabinet', 'soap', 'shoe_rack']
Task:Added meat to freezer
response:
| {
"thoughts": "After shopping for groceries, i returned home, i walked to the dining room and placed meat items in freezer located in the dining room. I opened up the bags from the grocery store, removed all the meat items, and after opening the freezer door, i placed the meat items into the freezer, and closed the door.",
"actions": [
"[WALK] <dining_room>",
"[WALK] <freezer>",
"[FIND] <freezer>",
"[OPEN] <freezer>",
"[FIND] <food_food>",
"[GRAB] <food_food>",
"[PUTBACK] <food_food> <freezer>",
"[CLOSE] <freezer>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['jelly', 'freezer', 'walllamp', 'bathroom', 'bills', 'needle', 'drawing', 'character', 'table_cloth', 'toaster', 'picture', 'floor_lamp', 'window', 'mousepad', 'microphone', 'food_kiwi', 'wall', 'wall_clock', 'newspaper', 'bed', 'television', 'tray', 'cupboard', 'controller', 'door', 'sink', 'bathtub', 'bathroom_cabinet', 'light', 'home_office']
Task:Turn on light
response:
{
"thoughts": I go to bedroom and locate a lamp. This lamp can be turned on by touch, as it has capacitive sensing. So I do it.,
"actions": ['[WALK] <bedroom>', '[WALK] <floor_lamp>', '[FIND] <floor_lamp>', '[TOUCH] <floor_lamp>', '[SWITCHON] <floor_lamp>']
}
|
Scene_Objects:['pot', 'tablelamp', 'microwave', 'toilet', 'chair', 'freezer', 'drawing', 'knifeblock', 'orchid', 'bathroom_counter', 'floor', 'bookshelf', 'keyboard', 'mat', 'shoe-shine_kit', 'window', 'foundation', 'cup', 'photoframe', 'check', 'coffe_maker', 'centerpiece', 'kitchen_counter', 'shoes', 'bathtub', 'sink', 'bathroom_cabinet', 'light', 'napkin', 'instrument_violin']
Task:Look at mirror
response:
| {
"thoughts": "I look at my reflection to check my cleanliness.",
"actions": [
"[WALK] <bathroom>",
"[WALK] <mirror>",
"[FIND] <mirror>",
"[TURNTO] <mirror>",
"[LOOKAT] <mirror>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['tablelamp', 'mop', 'walllamp', 'drawing', 'wallshelf', 'floor', 'ceiling', 'phone', 'desk', 'bookshelf', 'computer', 'toaster', 'mat', 'tvstand', 'ceilinglamp', 'couch', 'candle', 'pillow', 'wall', 'powersocket', 'photoframe', 'light_bulb', 'check', 'television', 'rag', 'doorjamb', 'kitchen_counter', 'door', 'bathroom_cabinet', 'light']
Task:Relax on sofa
response:
{
"thoughts": I go to the living room, switch off the light, sit in the sofa and do nothing.,
"actions": ['[WALK] <home_office>', '[WALK] <light>', '[FIND] <light>', '[SWITCHOFF] <light>', '[FIND] <couch>', '[SIT] <couch>', '[LIE] <couch>']
}
|
Scene_Objects:['filing_cabinet', 'tablelamp', 'toilet', 'chair', 'walllamp', 'detergent', 'trashcan', 'food_carrot', 'knifeblock', 'drawing', 'cpuscreen', 'knife', 'bathroom_counter', 'desk', 'keyboard', 'bookshelf', 'table_cloth', 'toaster', 'ceilinglamp', 'faucet', 'mousepad', 'pillow', 'dining_room', 'oven', 'photoframe', 'food_food', 'kitchen_counter', 'dresser', 'sink', 'home_office']
Task:Turn off light
response:
| {
"thoughts": "Walk to the lamp. Find the switch. Click the switch off.",
"actions": [
"[WALK] <bedroom>",
"[WALK] <floor_lamp>",
"[SWITCHOFF] <floor_lamp>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['sheets', 'microwave', 'chair', 'freezer', 'food_butter', 'drawing', 'food_egg', 'floor', 'ceiling', 'character', 'stovefan', 'ceilinglamp', 'faucet', 'shower', 'couch', 'wall', 'nightstand', 'bench', 'oven', 'cup', 'novel', 'ice', 'curtain', 'doorjamb', 'kettle', 'pencil', 'headset', 'sink', 'bathroom_cabinet', 'light']
Task:Drink
response:
{
"thoughts": Walk into kitchen, look for cup, walk to cup, look for refrigerator, walk to refrigerator, open door, grab water, pour water into cup,
"actions": ['[WALK] <dining_room>', '[WALK] <cup>', '[FIND] <cup>', '[TURNTO] <cup>', '[LOOKAT] <cup>', '[WALK] <cup>', '[FIND] <freezer>', '[TURNTO] <freezer>', '[LOOKAT] <freezer>', '[WALK] <freezer>', '[GRAB] <cup>', '[WALK] <water>', '[FIND] <water>', '[GRAB] <water>', '[POUR] <water> <cup>']
}
|
Scene_Objects:['pot', 'microwave', 'toilet', 'freezer', 'trashcan', 'drawing', 'wallshelf', 'cpuscreen', 'floor', 'desk', 'bookshelf', 'toaster', 'mat', 'ceilinglamp', 'plate', 'window', 'couch', 'pillow', 'wall', 'nightstand', 'bench', 'photoframe', 'coffe_maker', 'food_rice', 'dresser', 'cupboard', 'bathroom_cabinet', 'table', 'dishwasher', 'home_office']
Task:Set up table
response:
| {
"thoughts": "Get dishes. Place dishes neatly around table. Get food. Place food on middle of table.",
"actions": [
"[WALK] <dining_room>",
"[WALK] <plate>",
"[FIND] <plate>",
"[GRAB] <plate>",
"[WALK] <dining_room>",
"[WALK] <table>",
"[PUTBACK] <plate> <table>",
"[WALK] <dining_room>",
"[WALK] <plate>",
"[FIND] <plate>",
"[GRAB] <plate>",
"[WALK] <dining_room>",
"[WALK] <table>",
"[PUTBACK] <plate> <table>",
"[WALK] <dining_room>",
"[WALK] <plate>",
"[FIND] <plate>",
"[GRAB] <plate>",
"[WALK] <dining_room>",
"[WALK] <table>",
"[PUTBACK] <plate> <table>",
"[WALK] <dining_room>",
"[WALK] <food_food>",
"[FIND] <food_food>",
"[GRAB] <food_food>",
"[WALK] <dining_room>",
"[WALK] <table>",
"[PUTBACK] <food_food> <table>",
"[WALK] <dining_room>",
"[WALK] <food_food>",
"[FIND] <food_food>",
"[GRAB] <food_food>",
"[WALK] <dining_room>",
"[WALK] <table>",
"[PUTBACK] <food_food> <table>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['pot', 'filing_cabinet', 'bathroom', 'detergent', 'food_butter', 'drawing', 'wallshelf', 'floor', 'desk', 'bookshelf', 'plate', 'faucet', 'picture', 'iron', 'window', 'dish_soap', 'food_jam', 'wall', 'nightstand', 'toy', 'cutting_board', 'food_food', 'television', 'curtain', 'pajamas', 'conditioner', 'light', 'dishwasher', 'tooth_paste', 'shoe_rack']
Task:Wash dishes with dishwasher
response:
{
"thoughts": Go to the kitchen. Find the dishwasher. Open the dishwasher. Find the plates. Load the plates into the dishwasher. Find the soap. Load the soap into the dishwasher. Close the dishwasher. Turn on the dishwasher.,
"actions": ['[WALK] <dining_room>', '[WALK] <dishwasher>', '[FIND] <dishwasher>', '[OPEN] <dishwasher>', '[FIND] <plate>', '[GRAB] <plate>', '[PUTBACK] <plate> <dishwasher>', '[FIND] <plate>', '[GRAB] <plate>', '[PUTBACK] <plate> <dishwasher>', '[FIND] <dish_soap>', '[GRAB] <dish_soap>', '[PUTBACK] <dish_soap> <dishwasher>', '[CLOSE] <dishwasher>', '[SWITCHON] <dishwasher>']
}
|
Scene_Objects:['tablelamp', 'chair', 'bathroom', 'trashcan', 'orchid', 'drawing', 'cpuscreen', 'knifeblock', 'floor', 'dvd_player', 'computer', 'bookmark', 'brush', 'ceilinglamp', 'faucet', 'electric_shaver', 'floor_lamp', 'mousepad', 'wall', 'powersocket', 'novel', 'food_food', 'check', 'television', 'after_shave', 'cupboard', 'sink', 'home_office', 'table', 'tooth_paste']
Task:Turn on light
response:
| {
"thoughts": "Twist the lamp switch until the light goes on.",
"actions": [
"[WALK] <home_office>",
"[WALK] <floor_lamp>",
"[FIND] <floor_lamp>",
"[SWITCHON] <floor_lamp>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['walllamp', 'orchid', 'wallshelf', 'floor', 'ceiling', 'bedroom', 'bookshelf', 'mat', 'tvstand', 'ceilinglamp', 'shower', 'window', 'mousepad', 'dining_room', 'wall', 'nightstand', 'powersocket', 'photoframe', 'keys', 'television', 'curtain', 'food_onion', 'doorjamb', 'mouse', 'dresser', 'pajamas', 'sink', 'home_office', 'light', 'instrument_guitar']
Task:Turn off TV
response:
{
"thoughts": Walk over to tv and press power button,
"actions": ['[WALK] <home_office>', '[WALK] <television>', '[SWITCHOFF] <television>']
}
|
Scene_Objects:['fork', 'pot', 'sheets', 'chair', 'freezer', 'walllamp', 'milk', 'knife', 'drawing', 'kitchen_cabinet', 'floor', 'ceiling', 'toaster', 'mat', 'stovefan', 'window', 'food_jam', 'pillow', 'wall', 'bed', 'photoframe', 'food_food', 'check', 'curtain', 'towel_rack', 'mouse', 'kitchen_counter', 'cupboard', 'home_office', 'table']
Task:Look in refrigerator
response:
| {
"thoughts": "Approach the refrigerator. Use the handle to open the refrigerator door. Look at each shelf of the refrigerator. Look in each refrigerator drawer. Close the refrigerator door.",
"actions": [
"[WALK] <dining_room>",
"[WALK] <freezer>",
"[FIND] <freezer>",
"[TURNTO] <freezer>",
"[WALK] <freezer>",
"[OPEN] <freezer>",
"[TURNTO] <freezer>",
"[LOOKAT] <freezer>",
"[FIND] <kitchen_cabinet>",
"[OPEN] <kitchen_cabinet>",
"[FIND] <food_food>",
"[CLOSE] <kitchen_cabinet>",
"[CLOSE] <freezer>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['sheets', 'juice', 'tea', 'toilet', 'freezer', 'walllamp', 'dough', 'drawing', 'food_egg', 'floor', 'dvd_player', 'character', 'alcohol', 'stovefan', 'mat', 'shower', 'closetdrawer', 'clothes_hat', 'food_jam', 'wall', 'powersocket', 'curtain', 'doorjamb', 'mouse', 'kitchen_counter', 'laundry_detergent', 'door', 'light', 'dishwasher', 'table']
Task:Go to toilet
response:
{
"thoughts": I walk to the entrance of the bathroom. I turn to the door. I open the door to the bathroom. I walk inside of the bathroom. I find the toilet inside of the bathroom and walk over to it. I turn and look at the toilet.,
"actions": ['[WALK] <bathroom>', '[WALK] <toilet>', '[FIND] <toilet>', '[TURNTO] <toilet>', '[LOOKAT] <toilet>', '[WALK] <toilet>', '[TURNTO] <toilet>']
}
|
Scene_Objects:['sponge', 'pot', 'dog', 'walllamp', 'knifeblock', 'cpuscreen', 'floor', 'ceiling', 'phone', 'bedroom', 'keyboard', 'computer', 'stovefan', 'faucet', 'oil', 'pillow', 'wall', 'wall_clock', 'oven', 'pasta', 'keys', 'check', 'tray', 'food_onion', 'kitchen_counter', 'cupboard', 'door', 'sink', 'bathtub', 'light']
Task:Keep an eye on stove as something is cooking
response:
| {
"thoughts": "Watch whatever is cooking on the stove so it doesn't burn or pour. Turn off heat when done",
"actions": [
"[WALK] <dining_room>",
"[WALK] <oven>",
"[FIND] <pot>",
"[TURNTO] <pot>",
"[LOOKAT] <pot>",
"[FIND] <wall_clock>",
"[TURNTO] <wall_clock>",
"[LOOKAT] <wall_clock>",
"[SWITCHOFF] <oven>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['spectacles', 'pot', 'scissors', 'microwave', 'chair', 'freezer', 'drawing', 'orchid', 'cpuscreen', 'floor', 'ceiling', 'toaster', 'tvstand', 'faucet', 'dining_room', 'wall', 'bed', 'novel', 'food_food', 'mouthwash', 'rag', 'doorjamb', 'mouse', 'kitchen_counter', 'dresser', 'cupboard', 'home_office', 'bathroom_cabinet', 'soap', 'shoe_rack']
Task:Added meat to freezer
response:
{
"thoughts": After shopping for groceries, i returned home, i walked to the dining room and placed meat items in freezer located in the dining room. I opened up the bags from the grocery store, removed all the meat items, and after opening the freezer door, i placed the meat items into the freezer, and closed the door.,
"actions": ['[WALK] <dining_room>', '[WALK] <freezer>', '[FIND] <freezer>', '[OPEN] <freezer>', '[FIND] <food_food>', '[GRAB] <food_food>', '[PUTBACK] <food_food> <freezer>', '[CLOSE] <freezer>']
}
|
Scene_Objects:['filing_cabinet', 'tablelamp', 'microwave', 'toilet', 'freezer', 'knifeblock', 'drawing', 'wallshelf', 'ceiling', 'phone', 'desk', 'keyboard', 'face', 'toaster', 'laptop', 'tvstand', 'shower', 'stereo', 'homework', 'oil', 'towel', 'oven', 'basket_for_clothes', 'tray', 'towel_rack', 'hair', 'cupboard', 'door', 'bathtub', 'clothes_shirt']
Task:Rain welcome
response:
| {
"thoughts": "The robot would greet me and then hand me a towel. It would make an appropriate quip about the weather while I dried off. It would receive the wet towel when I was done.",
"actions": [
"[WALK] <bedroom>",
"[WALK] <towel>",
"[FIND] <towel>",
"[GRAB] <towel>",
"[FIND] <face>",
"[WIPE] <face>",
"[FIND] <hair>",
"[WIPE] <hair>",
"[PUTOBJBACK] <towel>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['food_bread', 'tablelamp', 'chair', 'freezer', 'drawing', 'orchid', 'needle', 'floor', 'phone', 'mail', 'mat', 'plate', 'faucet', 'stereo', 'couch', 'dustpan', 'dining_room', 'wall', 'bench', 'oven', 'toy', 'cup', 'newspaper', 'standingmirror', 'food_food', 'razor', 'console', 'towel_rack', 'bathtub', 'table']
Task:Set up table
response:
{
"thoughts": Get dishes. Place dishes neatly around table. Get food. Place food on middle of table.,
"actions": ['[WALK] <dining_room>', '[WALK] <plate>', '[FIND] <plate>', '[GRAB] <plate>', '[WALK] <dining_room>', '[WALK] <table>', '[PUTBACK] <plate> <table>', '[WALK] <dining_room>', '[WALK] <plate>', '[FIND] <plate>', '[GRAB] <plate>', '[WALK] <dining_room>', '[WALK] <table>', '[PUTBACK] <plate> <table>', '[WALK] <dining_room>', '[WALK] <plate>', '[FIND] <plate>', '[GRAB] <plate>', '[WALK] <dining_room>', '[WALK] <table>', '[PUTBACK] <plate> <table>', '[WALK] <dining_room>', '[WALK] <food_food>', '[FIND] <food_food>', '[GRAB] <food_food>', '[WALK] <dining_room>', '[WALK] <table>', '[PUTBACK] <food_food> <table>', '[WALK] <dining_room>', '[WALK] <food_food>', '[FIND] <food_food>', '[GRAB] <food_food>', '[WALK] <dining_room>', '[WALK] <table>', '[PUTBACK] <food_food> <table>']
}
|
Scene_Objects:['glue', 'filing_cabinet', 'tablelamp', 'chair', 'bathroom', 'drawing', 'wallshelf', 'clothes_underwear', 'hanger', 'character', 'bookshelf', 'clothes_jacket', 'piano_bench', 'ceilinglamp', 'shower', 'mousepad', 'teeth', 'pillow', 'check', 'toothbrush_holder', 'coin', 'doorjamb', 'mouse', 'kitchen_counter', 'bathtub', 'sink', 'bathroom_cabinet', 'table', 'dishwasher', 'coffee']
Task:Close door
response:
| {
"thoughts": "Push the door shut until it latches.",
"actions": [
"[RUN] <home_office>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['tablelamp', 'jelly', 'drawing', 'floor', 'bedroom', 'table_cloth', 'stovefan', 'mat', 'plate', 'window', 'couch', 'wall', 'nightstand', 'wall_clock', 'cutting_board', 'bed', 'novel', 'tray', 'curtain', 'doorjamb', 'diary', 'mouse', 'pencil', 'cd', 'clothes_dress', 'bathroom_cabinet', 'table', 'light', 'vase', 'laser_pointer']
Task:Read book
response:
{
"thoughts": I would go lie down in my bed and open the book and start reading.,
"actions": ['[WALK] <bedroom>', '[WALK] <bed>', '[FIND] <bed>', '[LIE] <bed>', '[FIND] <novel>', '[GRAB] <novel>', '[READ] <novel>']
}
|
Scene_Objects:['dry_pasta', 'pot', 'measuring_cup', 'microwave', 'chair', 'orchid', 'floor', 'ceiling', 'phone', 'desk', 'bedroom', 'cleaning_solution', 'keyboard', 'mop_bucket', 'stovefan', 'faucet', 'couch', 'clothes_pants', 'coffe_maker', 'mirror', 'curtain', 'rag', 'coffee_pot', 'towel_rack', 'kitchen_counter', 'dresser', 'cupboard', 'door', 'sink', 'dishwasher']
Task:Clean mirror
response:
| {
"thoughts": "Spray the mirror with a cleaner, take a dry rag and wipe the mirror down until its clean.",
"actions": [
"[WALK] <bathroom>",
"[WALK] <cleaning_solution>",
"[FIND] <cleaning_solution>",
"[GRAB] <cleaning_solution>",
"[FIND] <rag>",
"[GRAB] <rag>",
"[POUR] <cleaning_solution> <rag>",
"[PUTOBJBACK] <cleaning_solution>",
"[WALK] <mirror>",
"[WIPE] <mirror>",
"[PUTOBJBACK] <rag>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['pot', 'food_bread', 'microwave', 'chair', 'knifeblock', 'drawing', 'keyboard', 'button', 'stovefan', 'toaster', 'shower', 'window', 'dustpan', 'pillow', 'dining_room', 'nightstand', 'tape', 'novel', 'television', 'tray', 'curtain', 'after_shave', 'doorjamb', 'kitchen_counter', 'cupboard', 'clothes_dress', 'sink', 'bathroom_cabinet', 'light', 'dishwasher']
Task:Browse internet
response:
{
"thoughts": I pick up my phone. I touch it to navigate to the web browser. I can swipe and tap the screen to visit different websites. Then I read or watch the content within those websites.,
"actions": ['[WALK] <home_office>', '[WALK] <phone>', '[FIND] <phone>', '[GRAB] <phone>', '[TOUCH] <phone>', '[TURNTO] <phone>', '[LOOKAT] <phone>']
}
|
Scene_Objects:['tablelamp', 'food_salt', 'microwave', 'freezer', 'walllamp', 'bathroom', 'cpuscreen', 'bathroom_counter', 'ceiling', 'keyboard', 'computer', 'food_bacon', 'table_cloth', 'ceilinglamp', 'faucet', 'picture', 'clothes_socks', 'towel', 'dining_room', 'wall', 'pasta', 'bed', 'centerpiece', 'kitchen_counter', 'door', 'bathtub', 'home_office', 'light', 'table', 'colander']
Task:Clean screen
response:
| {
"thoughts": "Use a soft rag to wipe the screen until there are no marks on it.",
"actions": [
"[WALK] <home_office>",
"[WALK] <rag>",
"[FIND] <rag>",
"[GRAB] <rag>",
"[WALK] <computer>",
"[WIPE] <computer>",
"[PUTOBJBACK] <rag>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['chair', 'walllamp', 'blow_dryer', 'orchid', 'cpuscreen', 'floor', 'ceiling', 'desk', 'keyboard', 'thread', 'computer', 'mat', 'ceilinglamp', 'tvstand', 'mousepad', 'closetdrawer', 'couch', 'pillow', 'dining_room', 'wall', 'photoframe', 'television', 'doorjamb', 'mouse', 'dresser', 'door', 'sink', 'light', 'table', 'creditcard']
Task:Browse internet
response:
{
"thoughts": Walk to home office. Walk to computer. Switch on Computer.,
"actions": ['[WALK] <home_office>', '[WALK] <computer>', '[SWITCHON] <computer>']
}
|
Scene_Objects:['filing_cabinet', 'mop', 'chair', 'bathroom', 'food_carrot', 'woman', 'drawing', 'orchid', 'bathroom_counter', 'floor', 'ceiling', 'man', 'keyboard', 'mat', 'faucet', 'window', 'foundation', 'nightstand', 'powersocket', 'cup', 'coffe_maker', 'food_food', 'television', 'tray', 'towel_rack', 'cd', 'dresser', 'sink', 'clothes_scarf', 'light']
Task:Greet guests
response:
| {
"thoughts": "Stand next to the door. When the door opens from the outside, greet the guests.",
"actions": [
"[RUN] <home_office>",
"[WALK] <woman>",
"[FIND] <woman>",
"[TURNTO] <woman>",
"[LOOKAT] <woman>",
"[GREET] <woman>",
"[FIND] <man>",
"[TURNTO] <man>",
"[LOOKAT] <man>",
"[GREET] <man>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['filing_cabinet', 'food_pizza', 'tablelamp', 'food_ice_cream', 'crayon', 'shampoo', 'keyboard', 'mail', 'bookmark', 'stovefan', 'faucet', 'pillow', 'dining_room', 'broom', 'wall', 'love_seat', 'photoframe', 'novel', 'food_food', 'wine', 'tray', 'after_shave', 'shaving_cream', 'kettle', 'headset', 'pajamas', 'laundry_detergent', 'door', 'bathroom_cabinet', 'shoe_rack']
Task:Read book
response:
{
"thoughts": Obtain a book from the book shelf. Sit in the easy chair. Read the book. Mark the stopping point with a bookmark. Put the book back on the book shelf.,
"actions": ['[WALK] <home_office>', '[WALK] <novel>', '[FIND] <novel>', '[GRAB] <novel>', '[FIND] <love_seat>', '[SIT] <love_seat>', '[READ] <novel>', '[FIND] <bookmark>', '[GRAB] <bookmark>', '[PUTBACK] <bookmark> <novel>', '[PUTOBJBACK] <novel>']
}
|
Scene_Objects:['microwave', 'dog', 'freezer', 'walllamp', 'cat', 'knifeblock', 'bills', 'desk', 'bookshelf', 'keyboard', 'computer', 'ceilinglamp', 'faucet', 'clothes_hat', 'broom', 'powersocket', 'bench', 'coffe_maker', 'food_food', 'television', 'mirror', 'rag', 'lighter', 'towel_rack', 'mouse', 'dresser', 'sink', 'table', 'clothes_skirt', 'shoe_rack']
Task:Clean mirror
response:
| {
"thoughts": "Spray glass cleaner on mirror. Wipe with a clean cloth.",
"actions": [
"[WALK] <bathroom>",
"[WALK] <cleaning_solution>",
"[FIND] <cleaning_solution>",
"[GRAB] <cleaning_solution>",
"[WALK] <mirror>",
"[PUTOBJBACK] <cleaning_solution>",
"[FIND] <rag>",
"[GRAB] <rag>",
"[WALK] <mirror>",
"[WIPE] <mirror>",
"[WIPE] <mirror>",
"[WIPE] <mirror>",
"[PUTOBJBACK] <rag>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['measuring_cup', 'microwave', 'chair', 'walllamp', 'cat', 'wallshelf', 'floor', 'character', 'desk', 'tea_bag', 'toaster', 'plate', 'couch', 'dining_room', 'box', 'newspaper', 'bed', 'toilet_paper', 'curtain', 'coffee_pot', 'mouse', 'kettle', 'cupboard', 'controller', 'clothes_dress', 'sink', 'bathroom_cabinet', 'table', 'dishwasher', 'soap']
Task:Pet cat
response:
{
"thoughts": Find Cat in living room. Walk to cat. Pet cat.,
"actions": ['[WALK] <home_office>', '[WALK] <cat>', '[FIND] <cat>', '[TURNTO] <cat>', '[LOOKAT] <cat>', '[TOUCH] <cat>']
}
|
Scene_Objects:['fork', 'pot', 'tablelamp', 'chair', 'walllamp', 'trashcan', 'knifeblock', 'bathroom_counter', 'keyboard', 'computer', 'brush', 'tvstand', 'shower', 'faucet', 'mousepad', 'wall', 'bench', 'cup', 'bed', 'photoframe', 'clothes_gloves', 'tray', 'wooden_spoon', 'towel_rack', 'dresser', 'cupboard', 'door', 'bathtub', 'napkin', 'bathroom_cabinet']
Task:Wash monitor
response:
| {
"thoughts": "Clean the grime off of my monitor screen when not in use",
"actions": [
"[WALK] <home_office>",
"[WALK] <computer>",
"[FIND] <computer>",
"[TURNTO] <computer>",
"[LOOKAT] <computer>",
"[SWITCHOFF] <computer>",
"[FIND] <brush>",
"[TURNTO] <brush>",
"[LOOKAT] <brush>",
"[GRAB] <brush>",
"[WIPE] <computer>",
"[DROP] <brush>",
"[FIND] <napkin>",
"[GRAB] <napkin>",
"[WASH] <napkin>",
"[WIPE] <computer>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['tablelamp', 'food_salt', 'toilet', 'chair', 'bathroom', 'cat', 'trashcan', 'drawing', 'orchid', 'wallshelf', 'clothes_underwear', 'phone', 'ceiling', 'dvd_player', 'desk', 'stovefan', 'faucet', 'shower', 'closetdrawer', 'dining_room', 'wall', 'wall_clock', 'powersocket', 'oven', 'photoframe', 'bed', 'tray', 'mouse', 'dresser', 'home_office']
Task:Keep cats out of room
response:
{
"thoughts": Stand in the doorway and create a loud noise so the cats will stay out.,
"actions": ['[WALK] <bedroom>', '[WALK] <cat>', '[FIND] <cat>', '[TURNTO] <cat>', '[LOOKAT] <cat>', '[FIND] <cat>', '[TURNTO] <cat>', '[LOOKAT] <cat>']
}
|
Scene_Objects:['sponge', 'pot', 'filing_cabinet', 'tablelamp', 'microwave', 'walllamp', 'knifeblock', 'wallshelf', 'cpuscreen', 'floor', 'ceiling', 'character', 'mat', 'plate', 'faucet', 'ceilinglamp', 'dish_soap', 'mousepad', 'couch', 'dining_room', 'wall', 'oven', 'bench', 'basket_for_clothes', 'towel_rack', 'kitchen_counter', 'dishrack', 'cupboard', 'sink', 'light']
Task:Wash dirty dishes
response:
| {
"thoughts": "Wet the sponge, apply soap to sponge, scrub dishes, rinse dishes, place on dish rack.",
"actions": [
"[WALK] <dining_room>",
"[WALK] <sink>",
"[FIND] <sponge>",
"[GRAB] <sponge>",
"[FIND] <faucet>",
"[SWITCHON] <faucet>",
"[PUTBACK] <sponge> <sink>",
"[GRAB] <sponge>",
"[FIND] <dish_soap>",
"[GRAB] <dish_soap>",
"[POUR] <dish_soap> <sponge>",
"[PUTOBJBACK] <dish_soap>",
"[FIND] <plate>",
"[GRAB] <plate>",
"[SCRUB] <plate>",
"[RINSE] <plate>",
"[FIND] <dishrack>",
"[PUTBACK] <plate> <dishrack>",
"[FIND] <plate>",
"[GRAB] <plate>",
"[SCRUB] <plate>",
"[RINSE] <plate>",
"[PUTBACK] <plate> <dishrack>",
"[RINSE] <sponge>",
"[PUTOBJBACK] <sponge>",
"[SWITCHOFF] <faucet>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['pot', 'tablelamp', 'chair', 'walllamp', 'drawing', 'orchid', 'cpuscreen', 'wallshelf', 'hanger', 'floor', 'phone', 'ceiling', 'desk', 'keyboard', 'computer', 'mat', 'ceilinglamp', 'tvstand', 'mousepad', 'couch', 'pillow', 'wall', 'nightstand', 'oven', 'powersocket', 'keys', 'television', 'tray', 'doorjamb', 'towel_rack', 'mouse', 'dresser', 'door', 'light']
Task:Write an email
response:
{
"thoughts": I need to send an email to my friend so I go into my office and sit down at the desk. I turn on the computer and type the email.,
"actions": ['[WALK] <home_office>', '[WALK] <desk>', '[FIND] <desk>', '[TURNTO] <desk>', '[LOOKAT] <desk>', '[FIND] <chair>', '[SIT] <chair>', '[FIND] <computer>', '[SWITCHON] <computer>', '[FIND] <keyboard>', '[TYPE] <keyboard>']
}
|
Scene_Objects:['pot', 'filing_cabinet', 'food_salt', 'microwave', 'toilet', 'freezer', 'bathroom', 'trashcan', 'knife', 'cpuscreen', 'ceiling', 'bedroom', 'bookshelf', 'brush', 'stovefan', 'toaster', 'window', 'couch', 'wall_clock', 'powersocket', 'toy', 'bed', 'television', 'ice', 'curtain', 'mouse', 'dresser', 'controller', 'bathroom_cabinet', 'shoe_rack']
Task:Pick up toys
response:
| {
"thoughts": "Pickup toys from around the room. Organize toys back into their appropriate spaces on the shelves and into the toy boxes.",
"actions": [
"[WALK] <bedroom>",
"[WALK] <toy>",
"[FIND] <toy>",
"[GRAB] <toy>",
"[WALK] <filing_cabinet>",
"[OPEN] <filing_cabinet>",
"[PUTBACK] <toy> <filing_cabinet>",
"[CLOSE] <filing_cabinet>",
"[FIND] <toy>",
"[GRAB] <toy>",
"[WALK] <dresser>",
"[OPEN] <dresser>",
"[PUTBACK] <toy> <dresser>",
"[CLOSE] <dresser>",
"[FIND] <toy>",
"[GRAB] <toy>",
"[WALK] <filing_cabinet>",
"[OPEN] <filing_cabinet>",
"[PUTBACK] <toy> <filing_cabinet>",
"[CLOSE] <filing_cabinet>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['freezer', 'walllamp', 'bathroom', 'drawing', 'cpuscreen', 'floor', 'ceiling', 'character', 'remote_control', 'computer', 'mat', 'ceilinglamp', 'tvstand', 'window', 'electric_shaver', 'mousepad', 'couch', 'pillow', 'wall', 'nightstand', 'powersocket', 'photoframe', 'television', 'curtain', 'mouse', 'kitchen_counter', 'dresser', 'bathtub', 'table', 'light']
Task:Watch TV
response:
{
"thoughts": watching TV from sofa,
"actions": ['[WALK] <home_office>', '[WALK] <television>', '[FIND] <television>', '[PLUGIN] <television>', '[SWITCHON] <television>', '[FIND] <couch>', '[SIT] <couch>', '[FIND] <remote_control>', '[GRAB] <remote_control>', '[DROP] <remote_control>', '[TURNTO] <television>', '[LOOKAT] <television>']
}
|
Scene_Objects:['filing_cabinet', 'tea', 'chair', 'knifeblock', 'knife', 'cpuscreen', 'wallshelf', 'floor', 'phone', 'character', 'desk', 'bedroom', 'bookshelf', 'food_bacon', 'mat', 'shoe-shine_kit', 'food_jam', 'wall', 'powersocket', 'oven', 'photoframe', 'novel', 'mouthwash', 'coffee_filter', 'tray', 'curtain', 'vacuum_cleaner', 'dresser', 'bathtub', 'soap']
Task:Read to child
response:
| {
"thoughts": "Read a book out loud to a child",
"actions": [
"[WALK] <bedroom>",
"[WALK] <child>",
"[FIND] <child>",
"[GREET] <child>",
"[WALK] <filing_cabinet>",
"[OPEN] <filing_cabinet>",
"[FIND] <novel>",
"[GRAB] <novel>",
"[WALK] <chair>",
"[PULL] <chair>",
"[SIT] <chair>",
"[TURNTO] <novel>",
"[LOOKAT] <novel>",
"[TURNTO] <novel>",
"[POINTAT] <novel>",
"[TURNTO] <novel>",
"[POINTAT] <novel>",
"[READ] <novel>",
"[READ] <novel>",
"[READ] <novel>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['chair', 'walllamp', 'bathroom', 'orchid', 'knife', 'cpuscreen', 'floor', 'ceiling', 'phone', 'desk', 'keyboard', 'computer', 'button', 'table_cloth', 'ceilinglamp', 'mousepad', 'closetdrawer', 'dining_room', 'wall', 'powersocket', 'shelf', 'television', 'doorjamb', 'mouse', 'dresser', 'door', 'home_office', 'light', 'table', 'laser_pointer']
Task:Write an email
response:
{
"thoughts": Find the computer. Turn on the computer. Open the e-mail program. Create a new message. Write the message.,
"actions": ['[WALK] <home_office>', '[WALK] <computer>', '[FIND] <computer>', '[SWITCHON] <computer>']
}
|
Scene_Objects:['pot', 'tablelamp', 'chair', 'freezer', 'walllamp', 'milk', 'floor', 'keyboard', 'table_cloth', 'stovefan', 'mat', 'tvstand', 'faucet', 'window', 'couch', 'clothes_hat', 'pillow', 'dining_room', 'toy', 'basket_for_clothes', 'coffe_maker', 'video_game_controller', 'tray', 'doorjamb', 'kettle', 'dresser', 'cupboard', 'clothes_dress', 'sink', 'table']
Task:Pick up toys
response:
| {
"thoughts": "Bend down and pick up the toys off of the floor.",
"actions": [
"[WALK] <bedroom>",
"[WALK] <toy>",
"[FIND] <toy>",
"[GRAB] <toy>",
"[WALK] <dresser>",
"[OPEN] <dresser>",
"[PUTIN] <toy> <dresser>",
"[FIND] <toy>",
"[GRAB] <toy>",
"[WALK] <dresser>",
"[PUTIN] <toy> <dresser>",
"[FIND] <toy>",
"[WALK] <dresser>",
"[GRAB] <toy>",
"[PUTIN] <toy> <dresser>",
"[CLOSE] <dresser>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['filing_cabinet', 'chair', 'walllamp', 'knifeblock', 'orchid', 'cpuscreen', 'floor', 'ceiling', 'dvd_player', 'desk', 'keyboard', 'computer', 'bookmark', 'stovefan', 'ceilinglamp', 'mat', 'mousepad', 'oil', 'food_peanut_butter', 'broom', 'wall', 'novel', 'diary', 'mouse', 'cupboard', 'clothes_dress', 'door', 'table', 'light', 'dirt']
Task:Browse internet
response:
{
"thoughts": Go to office. Turn on computer. Pull out chair. Sit in chair. Use keyboard and mouse.,
"actions": ['[WALK] <home_office>', '[WALK] <computer>', '[FIND] <computer>', '[SWITCHON] <computer>', '[FIND] <chair>', '[PULL] <chair>', '[SIT] <chair>', '[TURNTO] <computer>', '[LOOKAT] <computer>', '[FIND] <keyboard>', '[TYPE] <keyboard>', '[FIND] <mouse>', '[GRAB] <mouse>']
}
|
Scene_Objects:['filing_cabinet', 'beer', 'crayon', 'freezer', 'walllamp', 'bathroom', 'detergent', 'wallshelf', 'floor', 'ceiling', 'purse', 'computer', 'mat', 'tvstand', 'shower', 'window', 'couch', 'food_peanut_butter', 'dining_room', 'wall', 'photoframe', 'bed', 'television', 'curtain', 'diary', 'cupboard', 'controller', 'bathroom_cabinet', 'table', 'dishwasher']
Task:Wipe down baseboards please
response:
| {
"thoughts": "Use a rag to get all the dust off the baseboards please.",
"actions": [
"[WALK] <home_office>",
"[WALK] <rag>",
"[FIND] <rag>",
"[GRAB] <rag>",
"[WALK] <wall>",
"[PUTOBJBACK] <rag>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['food_salt', 'chair', 'ground_coffee', 'food_butter', 'dough', 'drawing', 'floor', 'desk', 'bookmark', 'mat', 'faucet', 'floor_lamp', 'window', 'mousepad', 'couch', 'dustpan', 'pillow', 'wall', 'nightstand', 'oven', 'cup', 'novel', 'television', 'curtain', 'doorjamb', 'chessboard', 'controller', 'maindoor', 'home_office', 'table']
Task:Read book
response:
{
"thoughts": First I walk into the living room, I find the lamp and switch it on. Then I find a book, grab it, sit in the sofa, open the book and read it. When I'm finished, I find a bookmark, grab it and put it in the book. Then I close the book, put it back, and switch off the lamp.,
"actions": ['[WALK] <home_office>', '[WALK] <floor_lamp>', '[FIND] <floor_lamp>', '[SWITCHON] <floor_lamp>', '[FIND] <novel>', '[GRAB] <novel>', '[FIND] <couch>', '[SIT] <couch>', '[READ] <novel>', '[FIND] <bookmark>', '[GRAB] <bookmark>', '[PUTBACK] <bookmark> <novel>', '[PUTOBJBACK] <novel>', '[SWITCHOFF] <floor_lamp>']
}
|
Scene_Objects:['sponge', 'pot', 'microwave', 'walllamp', 'cat', 'food_carrot', 'knifeblock', 'floor', 'ceiling', 'toaster', 'faucet', 'window', 'dish_soap', 'dining_room', 'wall', 'oven', 'cup', 'drinking_glass', 'curtain', 'mouse', 'kitchen_counter', 'cd', 'dishrack', 'shoes', 'dresser', 'cupboard', 'sink', 'spoon', 'bathroom_cabinet', 'bathtub']
Task:Clean dishes
response:
| {
"thoughts": "Clean the dishes from the sink.",
"actions": [
"[WALK] <dining_room>",
"[WALK] <sink>",
"[FIND] <sponge>",
"[GRAB] <sponge>",
"[FIND] <dish_soap>",
"[GRAB] <dish_soap>",
"[POUR] <dish_soap> <sponge>",
"[PUTOBJBACK] <dish_soap>",
"[FIND] <faucet>",
"[SWITCHON] <faucet>",
"[FIND] <drinking_glass>",
"[GRAB] <drinking_glass>",
"[SCRUB] <drinking_glass>",
"[RINSE] <drinking_glass>",
"[FIND] <dishrack>",
"[PUTBACK] <drinking_glass> <dishrack>",
"[FIND] <cup>",
"[GRAB] <cup>",
"[SCRUB] <cup>",
"[RINSE] <cup>",
"[PUTBACK] <cup> <dishrack>",
"[RINSE] <sponge>",
"[PUTOBJBACK] <sponge>",
"[SWITCHOFF] <faucet>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['fork', 'pot', 'tablelamp', 'microwave', 'dog', 'walllamp', 'trashcan', 'knifeblock', 'knife', 'kitchen_cabinet', 'floor', 'ceiling', 'toaster', 'stovefan', 'plate', 'faucet', 'ceilinglamp', 'oil', 'wall', 'oven', 'coffe_maker', 'drinking_glass', 'tray', 'kitchen_counter', 'cupboard', 'sink', 'spoon', 'bag', 'dishwasher', 'colander']
Task:Unload dishwasher
response:
{
"thoughts": Open the dishwasher door. Take out all of the dishes and silverware. Put the dishes and silverware in their drawers and cabinets. Close the dishwasher door.,
"actions": ['[WALK] <dining_room>', '[WALK] <dishwasher>', '[FIND] <dishwasher>', '[OPEN] <dishwasher>', '[FIND] <drinking_glass>', '[GRAB] <drinking_glass>', '[FIND] <cupboard>', '[OPEN] <cupboard>', '[PUTBACK] <drinking_glass> <cupboard>', '[FIND] <fork>', '[GRAB] <fork>', '[FIND] <kitchen_cabinet>', '[OPEN] <kitchen_cabinet>', '[PUTIN] <fork> <kitchen_cabinet>', '[FIND] <knife>', '[GRAB] <knife>', '[PUTIN] <knife> <kitchen_cabinet>', '[FIND] <spoon>', '[GRAB] <spoon>', '[PUTIN] <spoon> <kitchen_cabinet>', '[FIND] <plate>', '[GRAB] <plate>', '[PUTIN] <plate> <kitchen_cabinet>', '[CLOSE] <kitchen_cabinet>', '[CLOSE] <dishwasher>']
}
|
Scene_Objects:['sheets', 'food_ice_cream', 'toilet', 'walllamp', 'dough', 'drawing', 'knifeblock', 'cpuscreen', 'clothes_underwear', 'floor', 'ceiling', 'character', 'bedroom', 'window', 'couch', 'pillow', 'wall', 'powersocket', 'basket_for_clothes', 'bed', 'food_food', 'television', 'doorjamb', 'mouse', 'pencil', 'dishrack', 'door', 'home_office', 'light', 'bathroom_cabinet']
Task:Turn off light
response:
| {
"thoughts": "Grasp light switch, flip light switch.",
"actions": [
"[WALK] <bedroom>",
"[WALK] <light>",
"[FIND] <light>",
"[TURNTO] <light>",
"[SWITCHOFF] <light>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['pot', 'food_bread', 'tablelamp', 'juice', 'cd_player', 'band-aids', 'walllamp', 'cpuscreen', 'bathroom_counter', 'ceiling', 'desk', 'stovefan', 'stereo', 'homework', 'food_orange', 'nightstand', 'bench', 'bed', 'newspaper', 'photoframe', 'coffe_maker', 'check', 'food_onion', 'curtain', 'doorjamb', 'dresser', 'door', 'sink', 'bathtub', 'light']
Task:Listen to music
response:
{
"thoughts": I walk in the living room, then I go the the stereo and the cd player and turn them on.,
"actions": ['[WALK] <home_office>', '[WALK] <stereo>', '[FIND] <stereo>', '[SWITCHON] <stereo>', '[FIND] <cd_player>', '[SWITCHON] <cd_player>']
}
|
Scene_Objects:['sponge', 'drawing', 'orchid', 'cpuscreen', 'floor', 'ceiling', 'shampoo', 'character', 'desk', 'bookmark', 'mat', 'mousepad', 'couch', 'food_orange', 'pillow', 'wall', 'wall_clock', 'bed', 'photoframe', 'coffe_maker', 'check', 'doorjamb', 'towel_rack', 'door', 'bathtub', 'home_office', 'table', 'light', 'instrument_guitar', 'scrabble']
Task:Read magazine
response:
| {
"thoughts": "Please sit on the couch and read a magazine",
"actions": [
"[WALK] <home_office>",
"[WALK] <newspaper>",
"[FIND] <newspaper>",
"[GRAB] <newspaper>",
"[WALK] <couch>",
"[SIT] <couch>",
"[TURNTO] <newspaper>",
"[LOOKAT] <newspaper>",
"[READ] <newspaper>",
"[READ] <newspaper>",
"[READ] <newspaper>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['food_oatmeal', 'microwave', 'freezer', 'bathroom', 'nail_polish', 'dough', 'food_carrot', 'orchid', 'drawing', 'knife', 'kitchen_cabinet', 'floor', 'ceiling', 'desk', 'brush', 'mat', 'couch', 'pillow', 'food_cereal', 'wall', 'food_vegetable', 'food_food', 'television', 'console', 'cards', 'kitchen_counter', 'cupboard', 'hair', 'door', 'light']
Task:Put groceries in Fridge
response:
{
"thoughts": Walk to kitchen, find groceries in cabinet and move them to fridge and close fridge.,
"actions": ['[WALK] <dining_room>', '[WALK] <kitchen_cabinet>', '[FIND] <kitchen_cabinet>', '[FIND] <food_vegetable>', '[GRAB] <food_vegetable>', '[FIND] <freezer>', '[OPEN] <freezer>', '[PUTIN] <food_vegetable> <freezer>', '[FIND] <food_carrot>', '[GRAB] <food_carrot>', '[PUTIN] <food_carrot> <freezer>', '[FIND] <food_cereal>', '[GRAB] <food_cereal>', '[PUTIN] <food_cereal> <freezer>', '[FIND] <food_oatmeal>', '[GRAB] <food_oatmeal>', '[PUTIN] <food_oatmeal> <freezer>', '[FIND] <food_food>', '[GRAB] <food_food>', '[PUTIN] <food_food> <freezer>', '[FIND] <freezer>', '[CLOSE] <freezer>']
}
|
Scene_Objects:['glue', 'pot', 'filing_cabinet', 'walllamp', 'orchid', 'bathroom_counter', 'floor', 'phone', 'character', 'desk', 'computer', 'brush', 'toaster', 'mat', 'shower', 'faucet', 'electric_shaver', 'mousepad', 'dining_room', 'nightstand', 'oven', 'toy', 'cup', 'food_onion', 'mouse', 'dresser', 'cupboard', 'door', 'bathroom_cabinet', 'dishwasher']
Task:Put toys away
response:
| {
"thoughts": "Pick up all the toys on the floor and put them in their correct storage bin or shelf.",
"actions": [
"[WALK] <bedroom>",
"[WALK] <toy>",
"[FIND] <toy>",
"[GRAB] <toy>",
"[WALK] <filing_cabinet>",
"[OPEN] <filing_cabinet>",
"[PUTBACK] <toy> <filing_cabinet>",
"[FIND] <toy>",
"[GRAB] <toy>",
"[WALK] <filing_cabinet>",
"[PUTBACK] <toy> <filing_cabinet>",
"[FIND] <toy>",
"[GRAB] <toy>",
"[WALK] <filing_cabinet>",
"[PUTBACK] <toy> <filing_cabinet>",
"[CLOSE] <filing_cabinet>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['tablelamp', 'microwave', 'freezer', 'bathroom', 'cpuscreen', 'desk', 'tea_bag', 'alcohol', 'tvstand', 'oven', 'tape', 'cutting_board', 'cup', 'clothes_gloves', 'novel', 'toilet_paper', 'television', 'coffee_filter', 'wooden_spoon', 'mouse', 'cd', 'controller', 'maindoor', 'sink', 'home_office', 'light', 'dishwasher', 'soap', 'dirt', 'shoe_rack']
Task:Listen to music
response:
{
"thoughts": Walk into living room. Turn on radio.,
"actions": ['[WALK] <home_office>', '[WALK] <stereo>', '[FIND] <stereo>', '[SWITCHON] <stereo>']
}
|
Scene_Objects:['pot', 'food_bread', 'juice', 'food_ice_cream', 'food_oatmeal', 'microwave', 'chair', 'freezer', 'walllamp', 'ground_coffee', 'milk', 'knifeblock', 'drawing', 'knife', 'floor', 'ceiling', 'stovefan', 'toaster', 'mat', 'faucet', 'ceilinglamp', 'bowl', 'wall', 'oven', 'bench', 'cup', 'coffe_maker', 'food_donut', 'tray', 'kitchen_counter', 'cupboard', 'sink', 'napkin', 'table', 'dishwasher']
Task:Cook some food
response:
| {
"thoughts": "Put oatmeal into a bowl. Add milk. Heat in a microwave. Put pod of coffee into a coffee maker. Activate. Pour juice into a glass.",
"actions": [
"[WALK] <dining_room>",
"[WALK] <cupboard>",
"[FIND] <cupboard>",
"[OPEN] <cupboard>",
"[FIND] <bowl>",
"[GRAB] <bowl>",
"[FIND] <food_oatmeal>",
"[GRAB] <food_oatmeal>",
"[PUTBACK] <food_oatmeal> <bowl>",
"[GRAB] <food_oatmeal>",
"[PUTOBJBACK] <food_oatmeal>",
"[CLOSE] <cupboard>",
"[WALK] <freezer>",
"[OPEN] <freezer>",
"[FIND] <milk>",
"[GRAB] <milk>",
"[POUR] <milk> <bowl>",
"[PUTOBJBACK] <milk>",
"[CLOSE] <freezer>",
"[FIND] <microwave>",
"[OPEN] <microwave>",
"[PUTIN] <bowl> <microwave>",
"[CLOSE] <microwave>",
"[SWITCHON] <microwave>",
"[FIND] <ground_coffee>",
"[GRAB] <ground_coffee>",
"[FIND] <coffe_maker>",
"[OPEN] <coffe_maker>",
"[PUTBACK] <ground_coffee> <coffe_maker>",
"[CLOSE] <coffe_maker>",
"[SWITCHON] <coffe_maker>",
"[FIND] <cup>",
"[GRAB] <cup>",
"[WALK] <freezer>",
"[OPEN] <freezer>",
"[FIND] <juice>",
"[GRAB] <juice>",
"[POUR] <juice> <cup>",
"[PUTOBJBACK] <juice>",
"[CLOSE] <freezer>",
"[FIND] <table>",
"[PUTBACK] <cup> <table>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['pot', 'detergent', 'ceilingfan', 'trashcan', 'drawing', 'knife', 'floor', 'ceiling', 'toaster', 'ceilinglamp', 'tvstand', 'balanceball', 'faucet', 'window', 'picture', 'couch', 'wall', 'pasta', 'novel', 'coffe_maker', 'television', 'curtain', 'doorjamb', 'kitchen_counter', 'dresser', 'maindoor', 'table', 'light', 'vase', 'tooth_paste']
Task:Watch TV
response:
{
"thoughts": I sit on my couch and watch TV in my living room,
"actions": ['[WALK] <home_office>', '[WALK] <television>', '[SWITCHON] <television>', '[TURNTO] <television>', '[LOOKAT] <television>', '[TOUCH] <television>', '[TURNTO] <television>', '[LOOKAT] <television>', '[TOUCH] <television>', '[TURNTO] <television>', '[LOOKAT] <television>', '[WALK] <couch>', '[SIT] <couch>', '[TURNTO] <television>', '[LOOKAT] <television>', '[WATCH] <television>', '[WATCH] <television>', '[WATCH] <television>']
}
|
Scene_Objects:['filing_cabinet', 'tablelamp', 'microwave', 'dough', 'drawing', 'wallshelf', 'floor', 'character', 'bedroom', 'purse', 'mat', 'ceilinglamp', 'couch', 'food_jam', 'dining_room', 'wall', 'nightstand', 'oven', 'bench', 'bed', 'food_food', 'coffe_maker', 'tray', 'towel_rack', 'mouse', 'headset', 'laundry_detergent', 'sink', 'home_office', 'table']
Task:Read book
response:
| {
"thoughts": "Climb into bed, then read the novel",
"actions": [
"[WALK] <bedroom>",
"[WALK] <novel>",
"[FIND] <novel>",
"[GRAB] <novel>",
"[WALK] <bed>",
"[LIE] <bed>",
"[READ] <novel>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['filing_cabinet', 'tablelamp', 'microwave', 'chair', 'freezer', 'detergent', 'floor', 'phone', 'ceiling', 'bedroom', 'keyboard', 'stovefan', 'shower', 'faucet', 'window', 'closetdrawer', 'wall', 'wall_clock', 'powersocket', 'bench', 'photoframe', 'coffe_maker', 'check', 'television', 'food_food', 'ice', 'rag', 'towel_rack', 'kitchen_counter', 'cupboard']
Task:Turn on light
response:
{
"thoughts": I go home and it's dark so I can not see anything,
"actions": ['[WALK] <home_office>', '[WALK] <floor_lamp>', '[FIND] <floor_lamp>', '[SWITCHON] <floor_lamp>']
}
|
Scene_Objects:['glue', 'filing_cabinet', 'crayon', 'mop', 'toilet', 'chair', 'cpuscreen', 'floor', 'phone', 'character', 'desk', 'bookshelf', 'stovefan', 'toaster', 'tvstand', 'faucet', 'stereo', 'electric_shaver', 'wall', 'oven', 'food_food', 'television', 'mirror', 'curtain', 'lighter', 'coffee_pot', 'cupboard', 'hairdryer', 'bathroom_cabinet', 'dishwasher']
Task:Shave
response:
| {
"thoughts": "Pick up razor. Push the button to turn it on. Move it in a linear motion against the beard growth. Make sure to get the bottom of your chin, your cheeks, and mustache area.",
"actions": [
"[WALK] <bathroom>",
"[WALK] <electric_shaver>",
"[FIND] <electric_shaver>",
"[GRAB] <electric_shaver>",
"[FIND] <mirror>",
"[TURNTO] <mirror>",
"[SWITCHON] <electric_shaver>",
"[SWITCHOFF] <electric_shaver>",
"[PUTOBJBACK] <electric_shaver>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['sheets', 'pot', 'chair', 'freezer', 'detergent', 'trashcan', 'dough', 'clothes_underwear', 'folder', 'stamp', 'bookshelf', 'clothes_jacket', 'table_cloth', 'mousepad', 'couch', 'food_jam', 'clothes_hat', 'pillow', 'nightstand', 'bench', 'oven', 'cutting_board', 'bed', 'food_food', 'check', 'doorjamb', 'console', 'cupboard', 'home_office', 'table']
Task:Pick up phone
response:
{
"thoughts": i go find the phone. i pick up the phone,
"actions": ['[WALK] <home_office>', '[WALK] <phone>', '[FIND] <phone>', '[GRAB] <phone>']
}
|
Scene_Objects:['tablelamp', 'chair', 'freezer', 'food_carrot', 'milk', 'knifeblock', 'floor', 'ceiling', 'phone', 'bedroom', 'purse', 'toaster', 'stovefan', 'shower', 'faucet', 'couch', 'form', 'pillow', 'wall', 'oven', 'food_food', 'oven_mitts', 'shaving_cream', 'kitchen_counter', 'cd', 'dresser', 'cupboard', 'bathtub', 'table', 'tooth_paste']
Task:Get glass of milk
response:
| {
"thoughts": "Get an empty glass. Take milk from refrigerator and open it. Pour milk into glass.",
"actions": [
"[WALK] <dining_room>",
"[WALK] <cup>",
"[FIND] <cup>",
"[GRAB] <cup>",
"[WALK] <freezer>",
"[OPEN] <freezer>",
"[FIND] <milk>",
"[GRAB] <milk>",
"[POUR] <milk> <cup>",
"[PUTOBJBACK] <milk>",
"[CLOSE] <freezer>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['food_salt', 'crayon', 'microwave', 'food_carrot', 'orchid', 'knifeblock', 'bills', 'floor', 'bedroom', 'stamp', 'envelope', 'clothes_jacket', 'bookmark', 'toaster', 'faucet', 'towel', 'dining_room', 'wall', 'coffe_maker', 'food_food', 'video_game_controller', 'tray', 'lighter', 'diary', 'kitchen_counter', 'cupboard', 'sink', 'soap', 'scrabble', 'hands_both']
Task:Wash hands
response:
{
"thoughts": I have had my lunch and walked towards the sink and applied soap to my hands and washed cleanly.,
"actions": ['[WALK] <dining_room>', '[WALK] <sink>', '[FIND] <soap>', '[GRAB] <soap>', '[FIND] <hands_both>', '[WASH] <hands_both>', '[FIND] <towel>', '[GRAB] <towel>', '[WIPE] <hands_both>']
}
|
Scene_Objects:['ironing_board', 'sheets', 'tablelamp', 'detergent', 'knifeblock', 'bathroom_counter', 'phone', 'desk', 'keyboard', 'computer', 'ceilinglamp', 'tvstand', 'iron', 'window', 'couch', 'pillow', 'bowl', 'nightstand', 'food_cheese', 'video_game_controller', 'tray', 'curtain', 'doorjamb', 'towel_rack', 'kitchen_counter', 'pajamas', 'bag', 'dishwasher', 'board_game', 'clothes_shirt']
Task:Iron shirt
response:
| {
"thoughts": "Set up an ironing board and an iron. Lay a shirt on board. Run iron over shirt until it is free of wrinkles.",
"actions": [
"[WALK] <bedroom>",
"[WALK] <ironing_board>",
"[FIND] <ironing_board>",
"[FIND] <clothes_shirt>",
"[GRAB] <clothes_shirt>",
"[PUTBACK] <clothes_shirt> <ironing_board>",
"[FIND] <iron>",
"[GRAB] <iron>",
"[PLUGIN] <iron>",
"[PLUGOUT] <iron>",
"[PUTOBJBACK] <iron>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['toilet', 'freezer', 'food_butter', 'knifeblock', 'bathroom_counter', 'floor', 'ceiling', 'phone', 'bookshelf', 'mop_bucket', 'ceilinglamp', 'tvstand', 'iron', 'window', 'mousepad', 'microphone', 'towel', 'dining_room', 'bowl', 'wall', 'powersocket', 'bench', 'cup', 'novel', 'television', 'coin', 'rag', 'doorjamb', 'headset', 'light']
Task:Turn on light
response:
{
"thoughts": I walk in to the living room. I find the light switch. I turn on the lights.,
"actions": ['[WALK] <home_office>', '[WALK] <light>', '[FIND] <light>', '[SWITCHON] <light>']
}
|
Scene_Objects:['dry_pasta', 'toilet', 'chair', 'freezer', 'walllamp', 'drawing', 'orchid', 'knifeblock', 'wallshelf', 'cpuscreen', 'floor', 'ceiling', 'desk', 'tea_bag', 'computer', 'deck_of_cards', 'mousepad', 'couch', 'pillow', 'wall', 'child', 'bed', 'coin', 'coffee_filter', 'doorjamb', 'centerpiece', 'door', 'table', 'light', 'dishwasher']
Task:Let baby learn how to walk
response:
| {
"thoughts": "The baby tries to stand up and balances on the couch and there are no sharp edges so it's safe.",
"actions": [
"[WALK] <home_office>",
"[WALK] <child>",
"[FIND] <child>",
"[TOUCH] <child>",
"[GREET] <child>",
"[FIND] <couch>",
"[TURNTO] <couch>",
"[POINTAT] <couch>",
"[TURNTO] <child>",
"[LOOKAT] <child>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['blow_dryer', 'bathroom', 'food_carrot', 'bathroom_counter', 'floor', 'ceiling', 'desk', 'stamp', 'bedroom', 'food_cake', 'computer', 'shower', 'picture', 'oven', 'basket_for_clothes', 'photoframe', 'food_food', 'check', 'towel_rack', 'mouse', 'pencil', 'shoes', 'comb', 'hairbrush', 'cupboard', 'laundry_detergent', 'sink', 'napkin', 'light', 'washing_machine']
Task:Wash clothes
response:
{
"thoughts": Pick up my dirty clothes, load the washer, add detergent, turn on washer,
"actions": ['[WALK] <bedroom>', '[WALK] <basket_for_clothes>', '[GRAB] <basket_for_clothes>', '[WALK] <home_office>', '[WALK] <washing_machine>', '[OPEN] <washing_machine>', '[PUTBACK] <basket_for_clothes> <washing_machine>', '[FIND] <laundry_detergent>', '[GRAB] <laundry_detergent>', '[PUTBACK] <laundry_detergent> <washing_machine>', '[RUN] <washing_machine>']
}
|
Scene_Objects:['sheets', 'sponge', 'tablelamp', 'mop', 'bathroom', 'drawing', 'orchid', 'knifeblock', 'wallshelf', 'floor', 'ceiling', 'stamp', 'wine_glass', 'clothes_socks', 'couch', 'pillow', 'wall', 'nightstand', 'food_cheese', 'bed', 'photoframe', 'food_food', 'check', 'doorjamb', 'door', 'sink', 'bag', 'bathroom_cabinet', 'table', 'light']
Task:Entertain
response:
| {
"thoughts": "Give guests wine, cheese, crackers and sit and talk",
"actions": [
"[WALK] <home_office>",
"[WALK] <food_cheese>",
"[FIND] <food_cheese>",
"[FIND] <food_food>",
"[FIND] <couch>",
"[SIT] <couch>",
"[FIND] <wine_glass>",
"[GRAB] <wine_glass>",
"[DRINK] <wine_glass>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['pot', 'tablelamp', 'walllamp', 'orchid', 'floor', 'ceiling', 'character', 'desk', 'table_cloth', 'toaster', 'ceilinglamp', 'plate', 'faucet', 'stovefan', 'dish_soap', 'mousepad', 'couch', 'wall', 'powersocket', 'bed', 'coffe_maker', 'food_food', 'kitchen_counter', 'cupboard', 'controller', 'sink', 'dishwasher', 'soap', 'clothes_shirt', 'teeth']
Task:Wash dishes with dishwasher
response:
{
"thoughts": Walk to the kitchen. Open the dishwasher. Grab plate and put it in the dishwasher. Grab dish soap and put it in the dishwasher. Switch on the dishwasher.,
"actions": ['[WALK] <dining_room>', '[WALK] <dishwasher>', '[FIND] <dishwasher>', '[OPEN] <dishwasher>', '[FIND] <plate>', '[GRAB] <plate>', '[PUTBACK] <plate> <dishwasher>', '[FIND] <dish_soap>', '[GRAB] <dish_soap>', '[PUTBACK] <dish_soap> <dishwasher>', '[CLOSE] <dishwasher>', '[SWITCHON] <dishwasher>']
}
|
Scene_Objects:['pot', 'microwave', 'walllamp', 'nail_polish', 'knifeblock', 'drawing', 'wallshelf', 'kitchen_cabinet', 'floor', 'ceiling', 'bathroom_counter', 'character', 'desk', 'faucet', 'stereo', 'shower', 'mousepad', 'couch', 'wall', 'oven', 'cup', 'photoframe', 'television', 'doorjamb', 'kitchen_counter', 'chessboard', 'cupboard', 'sink', 'table', 'dirt']
Task:Get glass of water
response:
| {
"thoughts": "Get a glass from a cabinet. Hold it under the sink and turn the cold water on. Fill the glass. Turn off the water.",
"actions": [
"[WALK] <dining_room>",
"[WALK] <kitchen_cabinet>",
"[FIND] <kitchen_cabinet>",
"[OPEN] <kitchen_cabinet>",
"[FIND] <cup>",
"[GRAB] <cup>",
"[CLOSE] <kitchen_cabinet>",
"[WALK] <sink>",
"[FIND] <faucet>",
"[SWITCHON] <faucet>",
"[PUTBACK] <cup> <sink>",
"[GRAB] <cup>",
"[SWITCHOFF] <faucet>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['filing_cabinet', 'measuring_cup', 'ground_coffee', 'walllamp', 'trashcan', 'drawing', 'wallshelf', 'hanger', 'floor', 'ceiling', 'keyboard', 'stovefan', 'toaster', 'mat', 'faucet', 'shower', 'wall', 'nightstand', 'oven', 'coffe_maker', 'coffee_filter', 'water', 'tray', 'kitchen_counter', 'cd', 'cupboard', 'controller', 'hairdryer', 'sink', 'home_office']
Task:Make coffee
response:
{
"thoughts": Walk to Kitchen and find coffee maker, find coffee filter and place it in coffee maker. Find ground coffee and water. put both in coffee maker. Close the coffee maker and switch it on.,
"actions": ['[WALK] <dining_room>', '[WALK] <coffe_maker>', '[FIND] <coffe_maker>', '[OPEN] <coffe_maker>', '[FIND] <coffee_filter>', '[GRAB] <coffee_filter>', '[PUTBACK] <coffee_filter> <coffe_maker>', '[FIND] <ground_coffee>', '[GRAB] <ground_coffee>', '[PUTBACK] <ground_coffee> <coffe_maker>', '[WALK] <water>', '[FIND] <water>', '[GRAB] <water>', '[POUR] <water> <coffe_maker>', '[CLOSE] <coffe_maker>', '[SWITCHON] <coffe_maker>']
}
|
Scene_Objects:['fork', 'filing_cabinet', 'blender', 'chair', 'freezer', 'walllamp', 'nail_polish', 'detergent', 'drawing', 'knife', 'wallshelf', 'bathroom_counter', 'floor', 'desk', 'computer', 'mat', 'ceilinglamp', 'plate', 'faucet', 'mousepad', 'candle', 'microphone', 'bowl', 'wall', 'nightstand', 'bench', 'bed', 'keys', 'food_food', 'table']
Task:Eat
response:
| {
"thoughts": "I get my prepared dinner set on the dining table, sit down on chair, pick up utensils, eat food",
"actions": [
"[WALK] <dining_room>",
"[WALK] <food_food>",
"[FIND] <food_food>",
"[GRAB] <food_food>",
"[WALK] <dining_room>",
"[WALK] <table>",
"[FIND] <table>",
"[PUTBACK] <food_food> <table>",
"[FIND] <chair>",
"[SIT] <chair>",
"[FIND] <fork>",
"[GRAB] <fork>",
"[FIND] <knife>",
"[GRAB] <knife>",
"[FIND] <plate>",
"[EAT] <plate>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['pot', 'tablelamp', 'tea', 'band-aids', 'walllamp', 'nail_polish', 'knifeblock', 'bills', 'cpuscreen', 'hanger', 'shampoo', 'phone', 'floor', 'bedroom', 'desk', 'toaster', 'stereo', 'clothes_socks', 'clothes_pants', 'clothes_hat', 'bowl', 'dining_room', 'bench', 'newspaper', 'clothes_gloves', 'coffe_maker', 'curtain', 'diary', 'hairbrush', 'bathroom_cabinet']
Task:Listen to music
response:
{
"thoughts": I pull up my online classes on the computer to look at what the assigned reading is. I get my textbook, notebook, and mechanical pencil. As I'm reading, I take notes over important points. I then use those notes to complete the assignment for the week. Frequently that involves writing a post on an online discussion board that answers a question related to things from the reading.,
"actions": ['[WALK] <home_office>', '[WALK] <stereo>', '[FIND] <stereo>', '[SWITCHON] <stereo>']
}
|
Scene_Objects:['sheets', 'pot', 'microwave', 'walllamp', 'drawing', 'knifeblock', 'cpuscreen', 'orchid', 'wallshelf', 'needle', 'bathroom_counter', 'floor', 'ceiling', 'phone', 'bookshelf', 'faucet', 'wall', 'oven', 'powersocket', 'bench', 'after_shave', 'curtain', 'centerpiece', 'pencil', 'kitchen_counter', 'cupboard', 'sink', 'bathtub', 'bathroom_cabinet', 'soap']
Task:Wipe down counter
response:
| {
"thoughts": "Clean counter",
"actions": [
"[WALK] <dining_room>",
"[WALK] <rag>",
"[FIND] <rag>",
"[GRAB] <rag>",
"[WALK] <sink>",
"[FIND] <faucet>",
"[SWITCHON] <faucet>",
"[RINSE] <rag>",
"[SQUEEZE] <rag>",
"[SWITCHOFF] <faucet>",
"[WALK] <kitchen_counter>",
"[WIPE] <kitchen_counter>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['pot', 'tablelamp', 'chair', 'walllamp', 'drawing', 'orchid', 'wallshelf', 'cpuscreen', 'hanger', 'floor', 'ceiling', 'phone', 'desk', 'bookshelf', 'keyboard', 'computer', 'mat', 'ceilinglamp', 'tvstand', 'mousepad', 'couch', 'pillow', 'wall', 'nightstand', 'powersocket', 'oven', 'check', 'television', 'tray', 'doorjamb', 'towel_rack', 'mouse', 'kitchen_counter', 'dresser', 'door', 'light']
Task:Work
response:
{
"thoughts": you go into your office, switch on the light, find chair at desk, sit in the chair, switch on the computer and work.,
"actions": ['[WALK] <home_office>', '[WALK] <light>', '[FIND] <light>', '[SWITCHON] <light>', '[FIND] <desk>', '[FIND] <chair>', '[SIT] <chair>', '[FIND] <computer>', '[SWITCHON] <computer>']
}
|
Scene_Objects:['scissors', 'walllamp', 'bathroom', 'knifeblock', 'wallshelf', 'floor', 'ceiling', 'bookshelf', 'keyboard', 'stovefan', 'ceilinglamp', 'tvstand', 'shower', 'plate', 'window', 'toaster', 'clothes_pants', 'broom', 'powersocket', 'bench', 'bed', 'check', 'food_food', 'coffe_maker', 'after_shave', 'dresser', 'hair', 'door', 'sink', 'clothes_shirt']
Task:Get dressed
response:
| {
"thoughts": "Get some clothes together that make an appropriate outfit. Put them on",
"actions": [
"[WALK] <bedroom>",
"[WALK] <dresser>",
"[FIND] <dresser>",
"[OPEN] <dresser>",
"[FIND] <clothes_pants>",
"[GRAB] <clothes_pants>",
"[PUTON] <clothes_pants>",
"[FIND] <clothes_shirt>",
"[GRAB] <clothes_shirt>",
"[PUTON] <clothes_shirt>",
"[CLOSE] <dresser>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['chair', 'drawing', 'orchid', 'wallshelf', 'floor', 'ceiling', 'desk', 'bookshelf', 'computer', 'stovefan', 'toaster', 'laptop', 'faucet', 'closetdrawer', 'couch', 'bowl', 'dining_room', 'wall', 'oven', 'bench', 'bed', 'coffe_maker', 'food_food', 'tray', 'doorjamb', 'kitchen_counter', 'dresser', 'cupboard', 'door', 'bathtub']
Task:Cook some food
response:
{
"thoughts": I switch on the stove and cook my groceries there.,
"actions": ['[WALK] <dining_room>', '[WALK] <oven>', '[FIND] <oven>', '[SWITCHON] <oven>', '[FIND] <bowl>', '[GRAB] <bowl>', '[PUTBACK] <bowl> <oven>', '[FIND] <food_food>', '[SWITCHOFF] <oven>']
}
|
Scene_Objects:['food_ice_cream', 'chair', 'walllamp', 'drawing', 'knifeblock', 'cpuscreen', 'bathroom_counter', 'ceiling', 'phone', 'desk', 'bookshelf', 'cleaning_solution', 'mat', 'tvstand', 'faucet', 'shower', 'couch', 'wall', 'nightstand', 'bench', 'toy', 'photoframe', 'mirror', 'tray', 'towel_rack', 'paper_towel', 'door', 'sink', 'clothes_scarf', 'bathroom_cabinet']
Task:Clean mirror
response:
| {
"thoughts": "First spray the mirrors with Windex or window type spray. Take paper towels and wipe until sparkly clean. If they have streaks, repeat the process.",
"actions": [
"[WALK] <bathroom>",
"[WALK] <bathroom_cabinet>",
"[OPEN] <bathroom_cabinet>",
"[FIND] <paper_towel>",
"[GRAB] <paper_towel>",
"[CLOSE] <bathroom_cabinet>",
"[WALK] <mirror>",
"[FIND] <cleaning_solution>",
"[RINSE] <cleaning_solution>",
"[WIPE] <mirror>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['sponge', 'pot', 'microwave', 'walllamp', 'knifeblock', 'food_egg', 'floor', 'ceiling', 'desk', 'bedroom', 'ceilinglamp', 'plate', 'faucet', 'window', 'dish_soap', 'mousepad', 'bowl', 'wall', 'oven', 'powersocket', 'check', 'mouthwash', 'after_shave', 'kitchen_counter', 'dishrack', 'dresser', 'cupboard', 'sink', 'bathtub', 'dishwasher']
Task:Wash dishes by hand
response:
{
"thoughts": I go to the kitchen and turn on the faucet. I get dish soap and put it into the sink. I grab a sponge and a bowl and start cleaning it. I put the bowl into the dishrack. I find a plate and clean it as well. I put it into the dishrack, turn off the faucet and put the sponge back.,
"actions": ['[WALK] <dining_room>', '[WALK] <sink>', '[FIND] <faucet>', '[SWITCHON] <faucet>', '[FIND] <dish_soap>', '[GRAB] <dish_soap>', '[FIND] <bowl>', '[GRAB] <bowl>', '[POUR] <dish_soap> <sink>', '[PUTOBJBACK] <dish_soap>', '[FIND] <sponge>', '[GRAB] <sponge>', '[FIND] <bowl>', '[SCRUB] <bowl>', '[RINSE] <bowl>', '[FIND] <dishrack>', '[PUTBACK] <bowl> <dishrack>', '[FIND] <plate>', '[GRAB] <plate>', '[SCRUB] <plate>', '[RINSE] <plate>', '[PUTBACK] <plate> <dishrack>', '[SWITCHOFF] <faucet>', '[PUTOBJBACK] <sponge>']
}
|
Scene_Objects:['pot', 'microwave', 'toilet', 'food_carrot', 'drawing', 'orchid', 'wallshelf', 'floor', 'ceiling', 'folder', 'tea_bag', 'mat', 'shower', 'window', 'mousepad', 'couch', 'pillow', 'wall', 'pasta', 'nightstand', 'photoframe', 'novel', 'food_food', 'tray', 'curtain', 'doorjamb', 'door', 'table', 'light', 'board_game']
Task:Read book
response:
| {
"thoughts": "Bring book, lie on the sofa, read book",
"actions": [
"[WALK] <home_office>",
"[WALK] <novel>",
"[GRAB] <novel>",
"[WALK] <couch>",
"[LIE] <couch>",
"[READ] <novel>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['measuring_cup', 'chair', 'walllamp', 'trashcan', 'orchid', 'drawing', 'knifeblock', 'hanger', 'bathroom_counter', 'ceiling', 'bedroom', 'bookshelf', 'clothes_jacket', 'faucet', 'couch', 'clothes_pants', 'form', 'wall', 'cutting_board', 'photoframe', 'check', 'food_food', 'curtain', 'doorjamb', 'towel_rack', 'clothes_dress', 'door', 'sink', 'bathroom_cabinet', 'washing_machine']
Task:Wash clothes
response:
{
"thoughts": Walk to bathroom, open cabinet and collect clothes and put them into washing machine and wash clothes,
"actions": ['[WALK] <bathroom>', '[WALK] <bathroom_cabinet>', '[FIND] <bathroom_cabinet>', '[OPEN] <bathroom_cabinet>', '[FIND] <clothes_dress>', '[GRAB] <clothes_dress>', '[FIND] <washing_machine>', '[OPEN] <washing_machine>', '[PUTBACK] <clothes_dress> <washing_machine>', '[FIND] <clothes_pants>', '[GRAB] <clothes_pants>', '[PUTBACK] <clothes_pants> <washing_machine>', '[FIND] <clothes_jacket>', '[GRAB] <clothes_jacket>', '[PUTBACK] <clothes_jacket> <washing_machine>', '[CLOSE] <washing_machine>', '[SWITCHON] <washing_machine>']
}
|
Scene_Objects:['sheets', 'pot', 'chair', 'freezer', 'walllamp', 'bathroom', 'drawing', 'ceiling', 'dvd_player', 'character', 'bookshelf', 'keyboard', 'brush', 'ceilinglamp', 'toaster', 'mat', 'window', 'nightstand', 'bench', 'photoframe', 'food_food', 'after_shave', 'diary', 'kettle', 'shoes', 'chessboard', 'dresser', 'door', 'bathroom_cabinet', 'colander']
Task:Dust
response:
| {
"thoughts": "Using soft brush, carefully dust both Indonesian wedding doll statues.",
"actions": [
"[WALK] <home_office>",
"[WALK] <dresser>",
"[OPEN] <dresser>",
"[FIND] <brush>",
"[GRAB] <brush>",
"[CLOSE] <dresser>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['chair', 'freezer', 'trashcan', 'food_carrot', 'orchid', 'floor', 'ceiling', 'thread', 'mail', 'button', 'food_bacon', 'bookmark', 'ceilinglamp', 'stereo', 'faucet', 'mousepad', 'wall', 'bench', 'cup', 'food_food', 'razor', 'television', 'food_donut', 'towel_rack', 'console', 'kitchen_counter', 'hairbrush', 'cupboard', 'sink', 'vase']
Task:Added meat to freezer
response:
{
"thoughts": After shopping for groceries, i returned home, i walked to the dining room and placed meat items in freezer located in the dining room. I opened up the bags from the grocery store, removed all the meat items, and after opening the freezer door, i placed the meat items into the freezer, and closed the door.,
"actions": ['[WALK] <dining_room>', '[WALK] <freezer>', '[FIND] <freezer>', '[OPEN] <freezer>', '[FIND] <food_food>', '[GRAB] <food_food>', '[PUTBACK] <food_food> <freezer>', '[CLOSE] <freezer>']
}
|
Scene_Objects:['microwave', 'drawing', 'bathroom_counter', 'floor', 'ceiling', 'tea_bag', 'stovefan', 'mat', 'window', 'electric_shaver', 'clothes_socks', 'couch', 'pillow', 'dining_room', 'wall', 'bench', 'oven', 'toy', 'bed', 'mouthwash', 'tray', 'doorjamb', 'mouse', 'pencil', 'chessboard', 'door', 'bathtub', 'instrument_guitar', 'light', 'soap']
Task:Turn on light
response:
| {
"thoughts": "Hit light switch",
"actions": [
"[WALK] <bedroom>",
"[WALK] <light>",
"[SWITCHON] <light>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['sheets', 'food_pizza', 'tablelamp', 'blender', 'scissors', 'freezer', 'walllamp', 'ground_coffee', 'orchid', 'character', 'bedroom', 'bookshelf', 'alcohol', 'button', 'toaster', 'stereo', 'homework', 'electric_shaver', 'mousepad', 'couch', 'bowl', 'toy', 'clothes_gloves', 'coffe_maker', 'kitchen_counter', 'conditioner', 'napkin', 'table', 'light', 'dishwasher']
Task:Listen to music
response:
{
"thoughts": Walk into living room, look for radio, grab on button,
"actions": ['[WALK] <home_office>', '[WALK] <stereo>', '[FIND] <stereo>', '[TURNTO] <stereo>', '[LOOKAT] <stereo>', '[FIND] <button>', '[GRAB] <button>']
}
|
Scene_Objects:['tablelamp', 'crayon', 'toilet', 'chair', 'bathroom', 'orchid', 'floor', 'ceiling', 'character', 'bedroom', 'desk', 'computer', 'tvstand', 'faucet', 'stereo', 'food_orange', 'photoframe', 'novel', 'television', 'curtain', 'doorjamb', 'diary', 'towel_rack', 'mouse', 'shoes', 'dresser', 'cupboard', 'sink', 'bathtub', 'home_office']
Task:Get dressed
response:
| {
"thoughts": "I get dressed in my bedroom in the mornings.",
"actions": [
"[WALK] <bedroom>",
"[WALK] <shoes>",
"[FIND] <shoes>",
"[GRAB] <shoes>",
"[PUTON] <shoes>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['fork', 'drawing', 'wallshelf', 'hanger', 'bathroom_counter', 'floor', 'ceiling', 'phone', 'character', 'desk', 'tea_bag', 'mat', 'closetdrawer', 'couch', 'pillow', 'wall', 'powersocket', 'toy', 'photoframe', 'check', 'mouthwash', 'food_food', 'food_donut', 'wine', 'curtain', 'doorjamb', 'console', 'dresser', 'door', 'light']
Task:Pick up toys
response:
{
"thoughts": Bend down and pick up the toys off of the floor.,
"actions": ['[WALK] <bedroom>', '[WALK] <toy>', '[FIND] <toy>', '[GRAB] <toy>', '[WALK] <dresser>', '[OPEN] <dresser>', '[PUTIN] <toy> <dresser>', '[FIND] <toy>', '[GRAB] <toy>', '[WALK] <dresser>', '[PUTIN] <toy> <dresser>', '[FIND] <toy>', '[WALK] <dresser>', '[GRAB] <toy>', '[PUTIN] <toy> <dresser>', '[CLOSE] <dresser>']
}
|
Scene_Objects:['pot', 'juice', 'measuring_cup', 'microwave', 'chair', 'freezer', 'band-aids', 'walllamp', 'detergent', 'knifeblock', 'drawing', 'cpuscreen', 'floor', 'ceiling', 'mop_bucket', 'toaster', 'stovefan', 'faucet', 'couch', 'wall', 'oven', 'cup', 'tray', 'food_onion', 'kitchen_counter', 'cupboard', 'food_snack', 'sink', 'napkin', 'home_office']
Task:Have snack
response:
| {
"thoughts": "Pour soda into a glass. Get pretzels out of bag. Eat and drink",
"actions": [
"[WALK] <dining_room>",
"[WALK] <cup>",
"[FIND] <cup>",
"[GRAB] <cup>",
"[WALK] <freezer>",
"[OPEN] <freezer>",
"[FIND] <juice>",
"[GRAB] <juice>",
"[POUR] <juice> <cup>",
"[PUTIN] <juice> <freezer>",
"[CLOSE] <freezer>",
"[WALK] <cupboard>",
"[OPEN] <cupboard>",
"[FIND] <food_snack>",
"[GRAB] <food_snack>",
"[CLOSE] <cupboard>",
"[DRINK] <cup>",
"[EAT] <food_snack>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['water_glass', 'pot', 'freezer', 'walllamp', 'knifeblock', 'cpuscreen', 'floor', 'ceiling', 'desk', 'keyboard', 'computer', 'stovefan', 'ceilinglamp', 'faucet', 'oil', 'mousepad', 'food_peanut_butter', 'wall', 'oven', 'powersocket', 'food_food', 'water', 'tray', 'doorjamb', 'mouse', 'kitchen_counter', 'cupboard', 'door', 'sink', 'light']
Task:Drink
response:
{
"thoughts": I need to go to Kitchen. Get a cup. Fill cup. Drink,
"actions": ['[WALK] <dining_room>', '[WALK] <cupboard>', '[FIND] <cupboard>', '[OPEN] <cupboard>', '[FIND] <water_glass>', '[GRAB] <water_glass>', '[WALK] <freezer>', '[WALK] <water>', '[FIND] <water>', '[GRAB] <water>', '[POUR] <water> <water_glass>', '[DRINK] <water_glass>']
}
|
Scene_Objects:['spectacles', 'sheets', 'pot', 'tablelamp', 'tea', 'microwave', 'bathroom', 'orchid', 'drawing', 'wallshelf', 'knifeblock', 'bathroom_counter', 'floor', 'remote_control', 'table_cloth', 'stovefan', 'mat', 'ceilinglamp', 'faucet', 'tvstand', 'wall', 'nightstand', 'bed', 'newspaper', 'photoframe', 'television', 'food_rice', 'console', 'pencil', 'bathroom_cabinet']
Task:Watch TV
response:
| {
"thoughts": "Watching tv with a remote is always fun; lay back, relax and enjoy;",
"actions": [
"[WALK] <bedroom>",
"[WALK] <television>",
"[FIND] <television>",
"[SWITCHON] <television>",
"[FIND] <remote_control>",
"[GRAB] <remote_control>",
"[FIND] <bed>",
"[LIE] <bed>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['chair', 'walllamp', 'drawing', 'wallshelf', 'clothes_underwear', 'hanger', 'floor', 'ceiling', 'desk', 'bedroom', 'remote_control', 'computer', 'mat', 'tvstand', 'ceilinglamp', 'shower', 'closetdrawer', 'couch', 'pillow', 'wall', 'photoframe', 'food_food', 'television', 'doorjamb', 'towel_rack', 'pencil', 'shoes', 'dresser', 'sink', 'instrument_violin']
Task:Watch TV
response:
{
"thoughts": I go to the the living room, sit down on the sofa, and then I switch on the TV with the TV remote,
"actions": ['[WALK] <home_office>', '[WALK] <remote_control>', '[FIND] <remote_control>', '[GRAB] <remote_control>', '[FIND] <couch>', '[SIT] <couch>', '[FIND] <television>', '[SWITCHON] <television>']
}
|
Scene_Objects:['sponge', 'tablelamp', 'chair', 'bathroom', 'trashcan', 'drawing', 'cpuscreen', 'wallshelf', 'floor', 'ceiling', 'character', 'bookshelf', 'mat', 'tvstand', 'stereo', 'dining_room', 'wall', 'nightstand', 'powersocket', 'bed', 'photoframe', 'video_game_controller', 'curtain', 'doorjamb', 'lighter', 'kitchen_counter', 'dresser', 'cupboard', 'bathroom_cabinet', 'instrument_guitar']
Task:Listen to music
response:
| {
"thoughts": "Put iPhone into speaker holder, open up the Pandora app, listen to music.",
"actions": [
"[WALK] <bedroom>",
"[WALK] <phone>",
"[FIND] <phone>",
"[GRAB] <phone>",
"[WALK] <stereo>",
"[PLUGIN] <phone>",
"[SWITCHON] <stereo>",
"[FIND] <bed>",
"[SIT] <bed>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['filing_cabinet', 'microwave', 'walllamp', 'detergent', 'cat', 'drawing', 'orchid', 'wallshelf', 'hanger', 'bathroom_counter', 'floor', 'desk', 'bookshelf', 'mat', 'tvstand', 'ceilinglamp', 'window', 'couch', 'dustpan', 'pillow', 'wall', 'wall_clock', 'television', 'tray', 'curtain', 'towel_rack', 'kitchen_counter', 'cupboard', 'table', 'light']
Task:Pet cat
response:
{
"thoughts": I see my cat on the couch so I walk over, sit down and pet the cat.,
"actions": ['[WALK] <home_office>', '[WALK] <cat>', '[FIND] <cat>', '[TURNTO] <cat>', '[LOOKAT] <cat>', '[FIND] <couch>', '[SIT] <couch>', '[TOUCH] <cat>']
}
|
Scene_Objects:['sheets', 'tablelamp', 'microwave', 'bathroom', 'detergent', 'wallshelf', 'hanger', 'floor', 'character', 'keyboard', 'envelope', 'clothes_jacket', 'stovefan', 'stereo', 'window', 'food_peanut_butter', 'nightstand', 'clothes_gloves', 'food_food', 'food_donut', 'after_shave', 'centerpiece', 'towel_rack', 'cupboard', 'door', 'clothes_scarf', 'sink', 'bathroom_cabinet', 'table', 'home_office']
Task:Put on coat
response:
| {
"thoughts": "I remove my scarf from a coat peg in the entrance hall. I wrap the scarf around my neck",
"actions": [
"[WALK] <home_office>",
"[WALK] <hanger>",
"[TURNTO] <hanger>",
"[LOOKAT] <hanger>",
"[FIND] <clothes_scarf>",
"[GRAB] <clothes_scarf>",
"[PUTON] <clothes_scarf>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['tablelamp', 'beer', 'toilet', 'freezer', 'walllamp', 'bathroom', 'drawing', 'wallshelf', 'ceiling', 'phone', 'character', 'keyboard', 'toaster', 'mat', 'floor_lamp', 'oil', 'pillow', 'dining_room', 'nightstand', 'wall_clock', 'bench', 'novel', 'television', 'curtain', 'doorjamb', 'dresser', 'door', 'sink', 'bathroom_cabinet', 'table']
Task:Read book
response:
{
"thoughts": Turn a light on and read a book in the bedroom.,
"actions": ['[WALK] <bedroom>', '[WALK] <floor_lamp>', '[FIND] <floor_lamp>', '[SWITCHON] <floor_lamp>', '[FIND] <novel>', '[GRAB] <novel>', '[READ] <novel>']
}
|
Scene_Objects:['fork', 'beer', 'microwave', 'toilet', 'walllamp', 'bathroom', 'trashcan', 'dough', 'orchid', 'dvd_player', 'keyboard', 'piano_bench', 'brush', 'faucet', 'pillow', 'dining_room', 'powersocket', 'newspaper', 'coffe_maker', 'television', 'wine', 'curtain', 'doorjamb', 'towel_rack', 'kitchen_counter', 'conditioner', 'clothes_scarf', 'bathtub', 'bathroom_cabinet', 'creditcard']
Task:Lock door
response:
| {
"thoughts": "Turn the lock.",
"actions": [
"[WALK] <bedroom>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['food_bread', 'freezer', 'walllamp', 'trashcan', 'food_carrot', 'cpuscreen', 'floor', 'ceiling', 'folder', 'bedroom', 'character', 'ceilinglamp', 'tvstand', 'faucet', 'dustpan', 'dining_room', 'wall', 'nightstand', 'oven', 'bench', 'photoframe', 'bed', 'food_food', 'television', 'food_donut', 'food_rice', 'doorjamb', 'shaving_cream', 'table', 'vase']
Task:Put groceries in Fridge
response:
{
"thoughts": walk to kitchen and seprate the groceries and arrange in fridge.,
"actions": ['[WALK] <dining_room>', '[WALK] <food_food>', '[FIND] <food_food>', '[GRAB] <food_food>', '[FIND] <freezer>', '[OPEN] <freezer>', '[PUTIN] <food_food> <freezer>']
}
|
Scene_Objects:['filing_cabinet', 'scissors', 'walllamp', 'drawing', 'orchid', 'wallshelf', 'bathroom_counter', 'floor', 'ceiling', 'bookshelf', 'remote_control', 'mat', 'tvstand', 'ceilinglamp', 'shower', 'window', 'couch', 'pillow', 'wall', 'powersocket', 'photoframe', 'television', 'curtain', 'doorjamb', 'centerpiece', 'kitchen_counter', 'door', 'table', 'light', 'creditcard']
Task:Watch TV
response:
| {
"thoughts": "I like to set on my sofa.Turn on the tv. And watch Perry Mason. Monday's through Friday's.",
"actions": [
"[WALK] <home_office>",
"[WALK] <couch>",
"[FIND] <couch>",
"[SIT] <couch>",
"[FIND] <television>",
"[SWITCHON] <television>",
"[FIND] <remote_control>",
"[GRAB] <remote_control>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['sheets', 'spectacles', 'food_salt', 'freezer', 'cpuscreen', 'floor', 'ceiling', 'shampoo', 'desk', 'keyboard', 'purse', 'computer', 'ceilinglamp', 'window', 'mousepad', 'food_noodles', 'wall', 'powersocket', 'clothes_gloves', 'food_food', 'toilet_paper', 'television', 'check', 'tray', 'doorjamb', 'cards', 'mouse', 'door', 'light', 'table']
Task:Put groceries in Fridge
response:
{
"thoughts": Open the fridge, put the groceries in the fridge and close it.,
"actions": ['[WALK] <dining_room>', '[WALK] <freezer>', '[FIND] <freezer>', '[OPEN] <freezer>', '[FIND] <food_food>', '[GRAB] <food_food>', '[PUTIN] <food_food> <freezer>', '[CLOSE] <freezer>']
}
|
Scene_Objects:['pot', 'microwave', 'walllamp', 'bathroom', 'orchid', 'drawing', 'wallshelf', 'cpuscreen', 'floor', 'ceiling', 'desk', 'bookshelf', 'mat', 'tvstand', 'ceilinglamp', 'shower', 'window', 'couch', 'pillow', 'wall', 'powersocket', 'oven', 'photoframe', 'bed', 'television', 'curtain', 'doorjamb', 'door', 'light', 'table']
Task:Watch TV
response:
| {
"thoughts": "Sit on the couch. Get comfortable. Switch on the television. Stare at the television.",
"actions": [
"[WALK] <home_office>",
"[WALK] <television>",
"[FIND] <television>",
"[SWITCHON] <television>",
"[FIND] <couch>",
"[SIT] <couch>",
"[TURNTO] <television>",
"[WATCH] <television>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['filing_cabinet', 'tablelamp', 'chair', 'freezer', 'drawing', 'orchid', 'floor', 'ceiling', 'desk', 'bedroom', 'bookshelf', 'computer', 'mat', 'shoe-shine_kit', 'oil', 'pillow', 'wall', 'pasta', 'powersocket', 'cup', 'food_food', 'video_game_controller', 'toothbrush_holder', 'television', 'doorjamb', 'kettle', 'door', 'light', 'clothes_skirt', 'coffee']
Task:Turn on light
response:
{
"thoughts": switching on light in an area,
"actions": ['[WALK] <home_office>', '[WALK] <light>', '[FIND] <light>', '[SWITCHON] <light>']
}
|
Scene_Objects:['beer', 'microwave', 'toilet', 'chair', 'freezer', 'wallshelf', 'floor', 'ceiling', 'character', 'bedroom', 'computer', 'toaster', 'ceilinglamp', 'window', 'pillow', 'wall', 'powersocket', 'oven', 'bench', 'food_cheese', 'bed', 'photoframe', 'coffe_maker', 'video_game_controller', 'ice', 'doorjamb', 'lighter', 'kitchen_counter', 'cupboard', 'home_office']
Task:Eat cheese
response:
| {
"thoughts": "Open fridge. Get cheese. Eat it.",
"actions": [
"[WALK] <dining_room>",
"[WALK] <freezer>",
"[OPEN] <freezer>",
"[FIND] <food_cheese>",
"[GRAB] <food_cheese>",
"[EAT] <food_cheese>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['pot', 'microwave', 'freezer', 'walllamp', 'drawing', 'bathroom_counter', 'floor', 'ceiling', 'phone', 'stovefan', 'mat', 'ceilinglamp', 'faucet', 'window', 'clothes_pants', 'pillow', 'wall', 'wall_clock', 'oven', 'nightstand', 'cutting_board', 'shelf', 'clothes_gloves', 'food_food', 'coffe_maker', 'wooden_spoon', 'tray', 'kitchen_counter', 'sink', 'bathroom_cabinet']
Task:Keep an eye on stove as something is cooking
response:
{
"thoughts": Watch whatever is cooking on the stove so it doesn't burn or pour. Turn off heat when done,
"actions": ['[WALK] <dining_room>', '[WALK] <oven>', '[FIND] <pot>', '[TURNTO] <pot>', '[LOOKAT] <pot>', '[FIND] <wall_clock>', '[TURNTO] <wall_clock>', '[LOOKAT] <wall_clock>', '[SWITCHOFF] <oven>']
}
|
Scene_Objects:['band-aids', 'trashcan', 'woman', 'drawing', 'orchid', 'floor', 'ceiling', 'man', 'envelope', 'tvstand', 'stereo', 'shoe-shine_kit', 'couch', 'pillow', 'dining_room', 'wall', 'bench', 'cup', 'food_food', 'coffe_maker', 'razor', 'tray', 'doorjamb', 'mouse', 'dresser', 'door', 'bathroom_cabinet', 'table', 'light', 'home_office']
Task:Hang with friends
response:
| {
"thoughts": "Sit with friends. Talk",
"actions": [
"[WALK] <home_office>",
"[WALK] <man>",
"[FIND] <man>",
"[TURNTO] <man>",
"[LOOKAT] <man>",
"[GREET] <man>",
"[FIND] <woman>",
"[TURNTO] <woman>",
"[LOOKAT] <woman>",
"[GREET] <woman>",
"[WALK] <couch>",
"[SIT] <couch>",
"[TURNTO] <man>",
"[LOOKAT] <man>",
"[GREET] <man>",
"[TURNTO] <woman>",
"[LOOKAT] <woman>",
"[GREET] <woman>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['chair', 'walllamp', 'orchid', 'drawing', 'wallshelf', 'floor', 'ceiling', 'bedroom', 'tea_bag', 'bookshelf', 'keyboard', 'brush', 'mat', 'tvstand', 'ceilinglamp', 'shower', 'window', 'stovefan', 'shoe-shine_kit', 'pillow', 'wall', 'powersocket', 'photoframe', 'novel', 'television', 'wooden_spoon', 'curtain', 'doorjamb', 'bathroom_cabinet', 'light']
Task:Watch TV
response:
{
"thoughts": Walk into the living room. Sit on the couch. Turn on TV.,
"actions": ['[WALK] <home_office>', '[WALK] <television>', '[FIND] <television>', '[SWITCHON] <television>']
}
|
Scene_Objects:['pot', 'filing_cabinet', 'microwave', 'detergent', 'drawing', 'knifeblock', 'food_egg', 'floor', 'computer', 'toaster', 'stovefan', 'tvstand', 'stereo', 'window', 'food_noodles', 'mousepad', 'couch', 'pillow', 'dining_room', 'nightstand', 'pasta', 'bed', 'oven_mitts', 'tray', 'curtain', 'lighter', 'towel_rack', 'shoes', 'cd', 'pajamas']
Task:Put clothes away
response:
| {
"thoughts": "I take clean clothes and either place them in dresser drawers or hang them in a closet.",
"actions": [
"[WALK] <bedroom>",
"[WALK] <filing_cabinet>",
"[OPEN] <filing_cabinet>",
"[CLOSE] <filing_cabinet>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['scissors', 'microwave', 'toilet', 'knife', 'bathroom_counter', 'floor', 'ceiling', 'mop_bucket', 'stovefan', 'mat', 'faucet', 'mousepad', 'towel', 'wall', 'wall_clock', 'oven', 'bench', 'cutting_board', 'shelf', 'food_rice', 'tray', 'after_shave', 'towel_rack', 'kitchen_counter', 'sink', 'bathroom_cabinet', 'soap', 'laser_pointer', 'dirt', 'hands_both']
Task:Hand washing
response:
{
"thoughts": Turn on the faucet. Soap up your hands. Then rinse under the running water of the faucet. Turn off faucet and dry hands.,
"actions": ['[WALK] <bathroom>', '[WALK] <sink>', '[FIND] <faucet>', '[SWITCHON] <faucet>', '[FIND] <soap>', '[GRAB] <soap>', '[FIND] <hands_both>', '[SCRUB] <hands_both>', '[PUTOBJBACK] <soap>', '[WASH] <hands_both>', '[SWITCHOFF] <faucet>', '[FIND] <towel>', '[GRAB] <towel>', '[WIPE] <hands_both>']
}
|
Scene_Objects:['pot', 'tablelamp', 'ground_coffee', 'walllamp', 'trashcan', 'knifeblock', 'hanger', 'floor', 'phone', 'character', 'bookshelf', 'stovefan', 'tvstand', 'faucet', 'shower', 'mousepad', 'pillow', 'wall', 'oven', 'cutting_board', 'television', 'food_rice', 'water', 'tray', 'kitchen_counter', 'cd', 'dresser', 'cupboard', 'bathroom_cabinet', 'home_office']
Task:Make coffee
response:
| {
"thoughts": "Get my coffee pot going",
"actions": [
"[WALK] <dining_room>",
"[WALK] <coffee_pot>",
"[FIND] <coffee_pot>",
"[GRAB] <coffee_pot>",
"[WALK] <water>",
"[FIND] <water>",
"[GRAB] <water>",
"[POUR] <water> <coffee_pot>",
"[FIND] <ground_coffee>",
"[GRAB] <ground_coffee>",
"[PUTBACK] <ground_coffee> <coffee_pot>",
"[GRAB] <ground_coffee>",
"[PUTOBJBACK] <ground_coffee>",
"[WALK] <oven>",
"[PUTBACK] <coffee_pot> <oven>",
"[SWITCHON] <oven>"
]
} |
You serve as an AI task planner.
1. Your task is to create a plan to achieve a goal by converting it into a sequence of actions. Each action should follow the format "[verb] <target1> <target2>", where 'verb' represents the action, and 'target1' and 'target2' are optional arguments.
2. You must describe your plan in natural language at the beginning. After that, you should list all the actions together.
Ensure that you response can be parsed by Python json.loads.
Here is an example plan to achieve a goal for reference:
Scene_Objects:['spectacles', 'pot', 'filing_cabinet', 'food_ice_cream', 'hanger', 'floor', 'ceiling', 'character', 'bookshelf', 'mat', 'tvstand', 'faucet', 'ceilinglamp', 'wall', 'powersocket', 'photoframe', 'clothes_gloves', 'bed', 'food_food', 'check', 'ice', 'doorjamb', 'pencil', 'kitchen_counter', 'laundry_detergent', 'door', 'napkin', 'home_office', 'light', 'dishwasher']
Task:Turn on light
response:
{
"thoughts": switching on the light,
"actions": ['[WALK] <bedroom>', '[WALK] <light>', '[FIND] <light>', '[SWITCHON] <light>']
}
|
Scene_Objects:['microwave', 'toilet', 'walllamp', 'bathroom', 'detergent', 'trashcan', 'woman', 'knifeblock', 'needle', 'cpuscreen', 'desk', 'man', 'keyboard', 'bedroom', 'computer', 'stovefan', 'mat', 'ceilinglamp', 'nightstand', 'oven', 'cup', 'food_food', 'check', 'doorjamb', 'mouse', 'kitchen_counter', 'bathtub', 'bathroom_cabinet', 'table', 'light']
Task:Greet guests
response:
| {
"thoughts": "Answer the door. Smile and give a warm hello, or other appropriate greeting. With newcomers to your home, offer your hand to be shook. With old friends it is appropriate to hug, or kiss one another on the cheek.",
"actions": [
"[WALK] <home_office>",
"[WALK] <man>",
"[FIND] <man>",
"[GREET] <man>",
"[FIND] <woman>",
"[GREET] <woman>"
]
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.