opd 开源项目推荐（36 个）

datawhalechina /agentic-rl-lab

复现并研究面向 LLM 智能体的强化学习算法，包括 GRPO、GSPO、DAPO、OPD、Search-R1、ReTool、ALFWorld 等。

chrisliu298 /awesome-on-policy-distillation

一份精选的论文、技术报告、框架和大型语言模型政策提炼（OPD）工具合集

nick7nlp /Awesome-LLM-On-Policy-Distillation

A curated collection of papers and resources on On-Policy Distillation for Large Language Models.

earthians /marley

开源、企业与现代健康信息系统

worldbench /DanceOPD

🔥 DanceOPD: On-Policy Generative Field Distillation

vra /Thinking-with-Visual-Primitives-pytorch

Unofficial PyTorch reproduction of DeepSeek's Thinking with Visual Primitives.

THU-BPM /RLCSD

Source code of paper "RLCSD: Reinforcement Learning with Contrastive On-Policy Self-Distillation"

naver-ai /opd2

Official implementation of On-Policy Delta Distillation (OPD2)

zhaoyingjun /Tiny-R2

Tiny-R2: A hybrid architecture integrating SWA, CSA, HCA, mHC, and DSMoE under the DeepSeek V4 design paradigm, enabling single-GPU OPD post-training.

quchangle1 /TurnSight

The implementation for our paper: TurnSight: Turn-Level Hindsight Self-Distillation for Tool-Integrated Reasoning.

buptyqx /H-OPD

H-OPD: Confidence Aware Heterogeneous Multi-Teacher Multimodal On-policy Distillation

Masoudjafaripour /Embodied_Distilled_SMLM

A repo for distilling a large teacher into a small vision-language model for efficient embodied spatial reasoning and action planning.

LiangrunFlora /STEP-OPD

Official implementation of the paper "STEP-OPD: Rethinking Output Targets and Internal Dynamics in On-Policy Distillation for Diffusion Models. "

AdityaSharmaHub /rogisetu

Simplifying Hospital OPD System by bridging the gap between patients and hospitals.

CandyGoose /OPD_web

Основы профессиональной деятельности (ОПД) Программная инженерия (ПИиКТ) Нейротехнологии и программирование ИТМО

worldbench /SCOPE

🔥 Learning When to Trust via Selective Context Preference Optimization

sylvain-wei /ReVuE

[UnderReview'26] On-Policy Visual Evidence Distillation

spring0722 /Frappe-Health

Open Source Health Information System. Get your Frappe Health instance ready within a few seconds and uncover the potentials of digitising your operations.

jamessvme /health

Open Source Health Information System. Get your Frappe Health instance ready within a few seconds and uncover the potentials of digitising your operations.

surabhipandey18 /DocPal

DocPal is an AI-powered patient record and symptom checker app built with Flask, MySQL, and a neural network model for disease prediction.