NVIDIA-NeMo / NVIDIA-NeMo/Gym

gymnasium_agent: Reasoning Included in Past Conversation History

Open Beginner friendly
#3,500 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

community-request
Dominant language
Python
Stars
1.2k
Forks
349
Avg merge
1d 23h
Merged PRs (30d)
324

Description

Describe the bug

  • Conversation history should not include past thoughts
  • The gymnasium agent app collects the models outputs in new_outputs L187 and passes it back into the request body L174.

Steps/Code to reproduce bug

  • Run with an LLM where thinking is turned on

Expected behavior

  • I expect the reasoning item to be omitted

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start in responses_api_agents/gymnasium_agent/app.py around lines 174 and 187. Run the gymnasium agent with an LLM that has thinking enabled, then trace how new_outputs is added to the request body. Done means reasoning items are omitted from past conversation history while other outputs remain available.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
ai, api
Issue type
Bug
Difficulty
2/5
Estimated time
1-3 hours
Activity status
Active
Clarity
Clearly specified
Newbie friendliness
84/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.