Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
A curated collection of papers, technical reports, frameworks, and tools for on-policy distillation (OPD) of large language models
| Date | Stars |
|---|---|
| 2026-07-31 | 591 |
| 2026-08-06 | 591 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
# ⚗️ Awesome On-Policy Distillation <p align="center"> <a href="https://awesome.re"><img src="https://img.shields.io/badge/Awesome-%E2%9A%97%EF%B8%8F_On--Policy_Distillation-000000?style=for-the-badge&labelColor=000000" alt="Awesome On-Policy Distillation"></a> </p> <p align="center"> <!-- entry-count-start --><a href="#contents"><img src="https://img.shields.io/badge/Entries-462-000000?style=for-the-badge&labelColor=000000" alt="Entries"></a><!-- entry-count-end --> <a href="https://github.com/chrisliu298/awesome-on-policy-distillation/stargazers"><img src="https://img.shields.io/github/stars/chrisliu298/awesome-on-policy-distillation?style=for-the-badge&logo=github&logoColor=white&label=Stars&labelColor=000000&color=000000" alt="GitHub Stars"></a> <a href="https://github.com/chrisliu298/awesome-on-policy-distillation/network/members"><img src="https://img.shields.io/github/forks/chrisliu298/awesome-on-policy-distillation?style=for-the-badge&logo=github&logoColor=white&label=Forks&labelColor=000000&color=000000" alt="GitHub Forks"></a> <a href="https://github.com/chrisliu298/awesome-on-policy-distillation/commits"><img src="https://img.shields.io/github/last-commit/chrisliu298/awesome-on-policy-distillation?style=for-the-badge&logo=github&logoColor=white&label=Last%20Commit&labelColor=000000&color=000000" alt="Last Commit"></a> </p> A curated collection of papers, technical reports, frameworks, and tools for on-policy distillation (OPD) of large language models. > **On-policy distillation** trains a student on samples from its own evolving policy, while a teacher (external, privileged, or self-conditioned) provides dense supervision on those same samples. On-policy distillation (OPD) trains a student on trajectories sampled from its own policy while a teacher scores the student-visited prefixes with dense token-level guidance. This on-policy data collection reduces the train-inference distribution gap that affects off-policy KD/SFT on fixed traces. Depending on the estimator, OPD looks like GKD on student rollouts or policy-gradient/RL with teacher-defined per-token KL/log-prob rewards, making the natural contrast sparse outcome-reward RL rather than RL as a whole. As of 2026, OPD is a standard post-training primitive at Alibaba (Qwen3), DeepSeek (V4), Xiaomi (MiMo), Zhipu (GLM-5), NVIDIA (Nemotron-Cascade 2), and others. **Shipping today?** Jump to [Frameworks and Implementations](#frameworks-and-implementations). **New to OPD?** Read [Start Here](#start-here). ## Contents - [Start Here](#start-here) - [Surveys and Essays](#surveys-and-essays) - [Core OPD Papers](#core-opd-papers) - [Foundations](#foundations) - [Gap-Bridging](#gap-bridging) - [Stability and Objective Design](#stability-and-objective-design) - [Self-Distillation](#self-distillation) - [Context and Experience Internalization](#context-and-experience-internalization) - [Efficiency, Systems, and Privacy Variants](#efficiency-systems-and-privacy-variants) - [Taxonomy](#taxonomy) - [Adjacent and Enabling Work](#adjacent-and-enabling-work) - [Domain Extensions](#domain-extensions) - [Agents and Tool-Use](#agents-and-tool-use) - [Multimodal and Vision-Language](#multimodal-and-vision-language) - [Speech and Audio](#speech-and-audio) - [Diffusion, Flow, and Generative Media](#diffusion-flow-and-generative-media) - [Embodied, Robotics, and Control](#embodied-robotics-and-control) - [Speculative Decoding](#speculative-decoding-draft-model-training) - [Technical Reports and Industrial Recipes](#technical-reports-and-industrial-recipes) - [Frameworks and Implementations](#frameworks-and-implementations) - [Acknowledgments](#acknowledgments) - [Contributing](#contributing) - [Citation](#citation) ## Start Here A fast path through the field: 1. **Survey.** [OPD Survey](https://arxiv.org/abs/2604.00626) — taxonomy, methods, and open problems in one place. 2. **Foundations.** [MiniLLM](https://arxiv.org/abs/2306.08543), [GKD](htt
Excerpt of 132,951 characters
Read on GitHub
chrisliu298 · University of California, Santa Cruz · United States
200
1
1
Would you bet a product on this? Bounded 0–100 and slow moving.
matched fp:2870b2d7f2d64719, topic:reinforcement-learning, topic:rl
matched fp:2870b2d7f2d64719, topic:knowledge-distillation
matched fp:2870b2d7f2d64719, topic:rlhf
matched fp:2870b2d7f2d64719, topic:awesome, topic:awesome-list