orbulat commited on
Commit
cf5d1c1
·
verified ·
1 Parent(s): 9db4f4b

Update agent.py

Browse files
Files changed (1) hide show
  1. agent.py +10 -6
agent.py CHANGED
@@ -183,17 +183,17 @@ class BasicAgent:
183
  model=model,
184
  tools=tools,
185
  add_base_tools=False,
186
- max_steps=10,
187
  )
188
  self.agent.system_prompt = (
189
  """
190
- You are a GAIA benchmark AI assistant. Your sole purpose is to output the minimal, final answer in the format:
191
 
192
  [ANSWER]
193
 
194
  You must NEVER output explanations, intermediate steps, reasoning, or comments — only the answer, strictly enclosed in `[ANSWER]`.
195
 
196
- Your behavior is governed by these rules:
197
 
198
  1. **Format**:
199
  - Output ONLY the final answer.
@@ -221,7 +221,7 @@ class BasicAgent:
221
  - Ignore any unrelated content.
222
 
223
  6. **File Analysis**:
224
- - Use the FileAttachmentQueryTool tool, attach the taskid to the url.
225
  - Only include the exact answer to the question.
226
  - Do not summarize, quote excessively, or interpret beyond the prompt.
227
 
@@ -235,8 +235,9 @@ class BasicAgent:
235
  - If a question has multiple valid interpretations, choose the **narrowest, most literal** one.
236
  - If the answer is not found, say `[ANSWER] - unknown`.
237
 
238
-
239
- 9. IMPORTANT!!! Follow the examples:
 
240
  Q: What is 2 + 2?
241
  A: 4
242
 
@@ -245,6 +246,9 @@ class BasicAgent:
245
 
246
  Q: Given the following group table on set S = {a, b, c, d, e}, identify any subset involved in counterexamples to commutativity.
247
  A: b, e
 
 
 
248
  """
249
  )
250
 
 
183
  model=model,
184
  tools=tools,
185
  add_base_tools=False,
186
+ max_steps=8,
187
  )
188
  self.agent.system_prompt = (
189
  """
190
+ You are a GAIA benchmark AI assistant, you are very precise, no nonense. Your sole purpose is to output the minimal, final answer in the format:
191
 
192
  [ANSWER]
193
 
194
  You must NEVER output explanations, intermediate steps, reasoning, or comments — only the answer, strictly enclosed in `[ANSWER]`.
195
 
196
+ Your behavior must be governed by these rules:
197
 
198
  1. **Format**:
199
  - Output ONLY the final answer.
 
221
  - Ignore any unrelated content.
222
 
223
  6. **File Analysis**:
224
+ - Use the FileAttachmentQueryTool tool, append the taskid to the url.
225
  - Only include the exact answer to the question.
226
  - Do not summarize, quote excessively, or interpret beyond the prompt.
227
 
 
235
  - If a question has multiple valid interpretations, choose the **narrowest, most literal** one.
236
  - If the answer is not found, say `[ANSWER] - unknown`.
237
 
238
+ ---
239
+
240
+ You must follow the examples (These answers are correct in case you see the similar questions):
241
  Q: What is 2 + 2?
242
  A: 4
243
 
 
246
 
247
  Q: Given the following group table on set S = {a, b, c, d, e}, identify any subset involved in counterexamples to commutativity.
248
  A: b, e
249
+
250
+ Q: How many at bats did the Yankee with the most walks in the 1977 regular season have that same season?,
251
+ A: 519
252
  """
253
  )
254