File size: 2,988 Bytes
213d0a7 d733484 213d0a7 1792a78 213d0a7 1792a78 213d0a7 f413c0e 213d0a7 f413c0e 213d0a7 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 |
# data
If you are looking for our intermediate labeling version, please refer to [mango-ttic/data-intermediate](https://huggingface.co/datasets/mango-ttic/data-intermediate)
Find more about us at [mango.ttic.edu](https://mango.ttic.edu)
## Folder Structure
Each folder inside `data` contains the cleaned up files used during LLM inference and results evaluations. Here is the tree structure from game `data/night` .
```bash
data/night/
├── night.actions.json # list of mentioned actions
├── night.all2all.jsonl # all simple paths between any 2 locations
├── night.all_pairs.jsonl # all connectivity between any 2 locations
├── night.edges.json # list of all edges
├── night.locations.json # list of all locations
└── night.walkthrough # enriched walkthrough exported from Jericho simulator
```
## Variations
### 70-step vs all-step version
In our paper, we benchmark using the first 70 steps of the walkthrough from each game. We also provide all-step versions of both `data` and `data-intermediate` collection.
* **70-step** `data-70steps.tar.zst`: contains the first 70 steps of each walkthrough. If the complete walkthrough is shorter than 70 steps, then all steps are used.
* **All-step** `data.tar.zst`: contains all steps of each walkthrough.
### Word-only & Word+ID
* **Word-only** `data.tar.zst`: Nodes are annotated by additional descriptive text to distinguish different locations with similar names.
* **Word + Object ID** `data-objid.tar.zst`: variation of the word-only version, where nodes are labeled using minimaly fixed names with object id from Jericho simulator.
* **Word + Random ID** `data-randid.tar.zst`: variation of the Jericho ID version, where the Jericho object id replaced with randomly generated integer.
We primarily rely on the **word-only** version as benchmark, yet providing word+ID version for diverse benchmark settings.
## How to use
We use `data.tar.zst` as an example here.
### 1. download from Huggingface
#### by directly download
You can selectively download certain variation of your choice.
#### by git
Make sure you have [git-lfs](https://git-lfs.com) installed
```bash
git lfs install
git clone https://huggingface.co/datasets/mango-ttic/data
# or, use hf-mirror if your connection to huggingface.co is slow
# git clone https://hf-mirror.com/datasets/mango-ttic/data
```
If you want to clone without large files - just their pointers
```bash
GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/datasets/mango-ttic/data
# or, use hf-mirror if your connection to huggingface.co is slow
# GIT_LFS_SKIP_SMUDGE=1 git clone https://hf-mirror.com/datasets/mango-ttic/data
```
### 2. decompress
Because some json files are huge, we use tar.zst to package the data efficiently.
silently decompress
```bash
tar -I 'zstd -d' -xf data.tar.zst
```
or, verbosely decompress
```bash
zstd -d -c data.tar.zst | tar -xvf -
```
|