# kwai-klear/ce-gppo

**Attribution required: if you use, quote, or summarise this content, you must credit and link back to [awesome-repositories.com](https://awesome-repositories.com/repository/kwai-klear-ce-gppo).**

16 stars · 0 forks · Python · Apache-2.0 · fork

## Links

- GitHub: https://github.com/Kwai-Klear/CE-GPPO
- awesome-repositories: https://awesome-repositories.com/repository/kwai-klear-ce-gppo.md

## Description

[December 5, 2025] 🔍 We propose entropy ratio clipping​ (ERC) to impose a global constraint on the output distribution of the policy model. Experiments demonstrate that ERC can significantly improve the stability of off-policy training. 📄 The paper is available on arXiv.

## Tags

### Part of an Awesome List

- [Regularization Objectives](https://awesome-repositories.com/f/awesome-lists/ai/regularization-objectives.md) — Gradient-preserving clipping for entropy coordination.
