[Feature] Implement SAPO algorithm as an alternative to hard clipping in GRPO/GSPO
Open
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 8.5k
- Forks
- 1.3k
- Avg merge
- 5h 36m
- Merged PRs (30d)
- 22
Description
Do you have any plans to add support for Soft Adaptive Policy Optimization (SAPO) - the new RL algorithm that replaces hard clipping with smooth, temperature-controlled gating - to the slime framework?
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Begin by reviewing the existing GRPO/GSPO implementation in slime and determining where hard clipping is applied. Define the SAPO temperature-controlled gating behavior and verify that SAPO can be selected as an alternative algorithm for both GRPO and GSPO.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100