Skip to content

Conversation

@extreme1228
Copy link
Contributor

Title: Revise Prompts to Comply with OpenAI API Policy

Description:

Background

Recent updates to the OpenAI API have introduced stricter content filtering policies, causing some of our existing prompts (e.g., those containing "thought" or similar keywords) to trigger the filter and result in rejected requests. This impacts the functionality of [task/os 、 task/alfworld] by preventing successful API calls.

Changes

This pull request modifies the prompts in the following ways:

  • Removed or rephrased prompts containing "thought" or other flagged terms that trigger OpenAI's content filter (e.g., changed "Provide your thought process" to "Explain your reasoning").
  • Updated files:
    • AgentBench/src/server/tasks/os_interaction/task.py
    • AgentBench/src/server/tasks/alfworld/prompts/*
    • AgentBench/src/server/tasks/alfworld/task.py
  • Ensured the revised prompts maintain their original intent and functionality.

Impact

  • Prevents API requests from being blocked by OpenAI’s filter mechanism.
  • Maintains the quality and clarity of generated responses.
  • No significant performance or behavioral changes expected beyond filter avoidance.

Testing

  • Tested revised prompts in the environment; no filter triggers observed.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants