R1-onevision, a visual language model capable of deep CoT reasoning.

Muon optimizer: +>30% sample efficiency with <3% wallclock overhead

Wan: Open and Advanced Large-Scale Video Generative Models

A library for advanced large language model reasoning

DeepGEMM: clean and efficient FP8 GEMM kernels with fine-grained scaling

FlashMLA: Efficient MLA Decoding Kernel for Hopper GPUs

EasyR1: An Efficient, Scalable, Multi-Modality RL Training Framework based on veRL

A CPU Realtime VLM in 500M. Surpassed Moondream2 and SmolVLM. Training from scratch with ease.

Solve Visual Understanding with Reinforced VLMs

[ICLR 2025] MoE++: Accelerating Mixture-of-Experts Methods with Zero-Computation Experts

Production-tested AI infrastructure tools for efficient AGI development and community-driven innovation

Official Repo for Open-Reasoner-Zero

Train transformer language models with reinforcement learning.

Extend OpenRLHF to support LMM RL training for reproduction of DeepSeek-R1 on multimodal tasks.

verl: Volcano Engine Reinforcement Learning for LLMs

Collect every awesome work about r1!

Reasoning in LLMs: Papers and Resources, including Chain-of-Thought, OpenAI o1, and DeepSeek-R1 🍓

LIMO: Less is More for Reasoning

A framework to study AI models in Reasoning, Alignment, and use of Memory (RAM).

A paper list of some recent works about Token Compress for Vit and VLM

Align Anything: Training All-modality Model with Feedback

minimal-cost for training 0.5B R1-Zero

✨✨ [ICLR 2025] MME-RealWorld: Could Your Multimodal LLM Challenge High-Resolution Real-World Scenarios that are Difficult for Humans?

A fork to add multimodal model training to open-r1

A simple framework for experimenting with Reinforcement Learning in Python.

RAGFlow is an open-source RAG (Retrieval-Augmented Generation) engine based on deep document understanding.

A collection of resources that investigate social agents.