reinforcement-learning

Free AI/ML course with 950+ Jupyter notebooks — Python, deep learning, LLMs, RAG, agents, prompt engineering, fine-tuning, MLOps

🏆Winning Project | ModelGate is a contract-aware AI control plane that ingests customer contracts, extracts SLA/privacy/routing constraints, and generates an OpenAI-compatible endpoint that automatically routes every request to the optimal model. Simple queries go to cheap models. Complex queries go to premium ones.

This repository collects awesome representative papers and resources for "From Pre-training to Post-training: A Survey on Time Series Foundation Models" (TechRxiv-26).

Lightweight replication study of DeepSeek-R1-Zero. Interesting findings include "No Aha Moment", "Longer CoT ≠ Accuracy", and "Language Mixing in Instruct Models".

Implementation of Multi-Game Decision Transformers in PyTorch

Implementation of GATO style Generalist Multimodal model capable of image, text, RL and Robotics tasks

A Continual Learning Framework for Production LLM Agents

Pure RL to post-train base models for social reasoning capabilities. Lightweight replication of DeepSeek-R1-Zero with Social IQa dataset.

Machine learning robotics engineer preparation material.

[ICML 2024] Code for the paper "Confronting Reward Overoptimization for Diffusion Models: A Perspective of Inductive and Primacy Biases"

PipelineLLM 是一个系统性的大语言模型（LLM）后训练学习项目，涵盖从监督微调（SFT）到偏好优化（DPO）、强化学习（RLHF/PPO/GRPO）再到持续学习（Continual Learning)的完整技术栈。

Panacea is a framework for building collaborative, intelligent multi agent AI systems. The framework provides a robust infrastructure for creating and managing multiple AI agents, and enables developers and organizations to build, deploy, and optimize AI agents that work well in dynamic, complex environments.

Official codebase for "Quantile Reward Policy Optimization: Alignment with Pointwise Regression and Exact Partition Functions" (Matrenok et al. 2025).

Interview prep for LLM post-training, ML, system design, and coding — organized by career path.

A lightweight post-training framework for LLMs and VLMs. 51 algorithms, 38 verified models. Scales with DeepSpeed, vLLM, and Ray.

A multi-purpose repository with Sentiment Analysis of Stocks news, and Reinforcement learning based bot to trade stocks.

[ICLR'26] SketchThinker-R1: Towards Efficient Sketch-Style Reasoning in Large Multimodal Models

Multi-node distributed LLM training framework

KONASH: Train knowledge agents that search, retrieve, and reason. Based on KARL (Databricks, 2026).

Implementation of ResFit, Residual Off-Policy RL for Finetuning Behavior Cloning Policies