Repository navigation
feat(distillation): implement OPSA in AReaL - #1768
Open
zahrayousefijamarani wants to merge 9 commits into
Open
zahrayousefijamarani wants to merge 9 commits into
zahrayousefijamarani wants to merge 9 commits into
Conversation
zahrayousefijamarani
requested review from
CormickKneey,
HwVanICI,
Le8r0nJames,
PrometheusComing,
fishcrap,
garrett4wade,
nuzant,
rchardx and
sitabulaixizawaluduo
as code owners
October 1, 2026 22:16
Collaborator
|
Please run |
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Implements OPSA (On-Policy Self-Adaptation) in AReaL for RL post-training of language models.
This PR integrates the OPSA training workflow with AReaL's existing rollout and training infrastructure. The implementation uses token-level policy information from on-policy generations to construct the OPSA training signal without requiring a separate teacher model.
Key components include:
OPSA-based RL training workflow
Integration with AReaL's rollout and policy optimization pipeline
Token-level policy information and entropy computation for OPSA
Self-adaptive advantage construction following the OPSA method
DAPO-Math-17k dataset preparation for training
AIME 2024 dataset preparation for evaluation
OPSA training configuration and example entry point
README documentation covering the method, dataset preparation, configuration, and usage
Type of Change
Checklist
Contributing Guide
pre-commit run --all-files)./docs/build_all.sh)main/review-prcommand/create-prBreaking Change Details (if applicable):
Additional Context
Need help? Check the
Contributing Guide
or ask in GitHub Discussions!