A compact implementation of SGLang, designed to demystify the complexities of modern LLM serving systems.

SkyRL: A Modular Full-stack RL Library for LLMs

Tree Search for LLM Agent Reinforcement Learning

Codes for the paper "BAPO: Stabilizing Off-Policy Reinforcement Learning for LLMs via Balanced Policy Optimization with Adaptive Clipping" by Zhiheng Xi et al.

Tongyi Deep Research, the Leading Open-source Deep Research Agent

The official code of ARPO & AEPO

QeRL enables RL for 32B LLMs on a single H100 GPU.

Bridge Megatron-Core to Hugging Face/Reinforcement Learning

Build memory-native AI agents with Memory OS — an open-source framework for long-term memory, retrieval, and adaptive learning in large language models. Agent Memory | Memory System | Memory Manage…

slime is an LLM post-training framework for RL Scaling.

A high-throughput and memory-efficient inference and serving engine for LLMs

Scalable toolkit for efficient model reinforcement

Async RL Training at Scale

Muon is an optimizer for hidden layers in neural networks

Unleashing the Power of Reinforcement Learning for Math and Code Reasoners

Parallel Scaling Law for Language Model — Beyond Parameter and Inference Time Scaling

Democratizing Reinforcement Learning for LLMs

Mooncake is the serving platform for Kimi, a leading LLM service provided by Moonshot AI.

FireFlyer Record file format, writer and reader for DL training samples.

Lightning-Fast RL for LLM Reasoning and Agents. Made Simple & Flexible.

Official Repo for Open-Reasoner-Zero

LoRA-One: One-Step Full Gradient Could Suffice for Fine-Tuning Large Language Models, Provably and Efficiently (ICML2025 Oral)

A Self-adaptation Framework🐙 that adapts LLMs for unseen tasks in real-time!

🤗 smolagents: a barebones library for agents that think in code.

Fine-tuning & Reinforcement Learning for LLMs. 🦥 Train OpenAI gpt-oss, DeepSeek, Qwen, Llama, Gemma, TTS 2x faster with 70% less VRAM.