EvolvingLMMs-Lab/open-r1-multimodal
A fork of open-r1 that adds multimodal RL training support for vision-language models using the GRPO algorithm.
Not currently ranked — collecting fresh signals.
star history
This repository extends the open-r1 project to support multimodal reasoning model training. It implements the GRPO (Group Relative Policy Optimization) algorithm for training vision-language models like Qwen2-VL and Aria-MoE on math reasoning tasks. The project provides open-sourced training datasets with reasoning paths and verifiable answers, trained model checkpoints, and scripts for creating custom multimodal reasoning data.
Frequently asked
- What is EvolvingLMMs-Lab/open-r1-multimodal?
- A fork of open-r1 that adds multimodal RL training support for vision-language models using the GRPO algorithm.
- Is open-r1-multimodal open source?
- Yes — EvolvingLMMs-Lab/open-r1-multimodal is open source, released under the Apache-2.0 license.
- What language is open-r1-multimodal written in?
- EvolvingLMMs-Lab/open-r1-multimodal is primarily written in Python.
- How popular is open-r1-multimodal?
- EvolvingLMMs-Lab/open-r1-multimodal has 1.6k stars on GitHub.
- Where can I find open-r1-multimodal?
- EvolvingLMMs-Lab/open-r1-multimodal is on GitHub at https://github.com/EvolvingLMMs-Lab/open-r1-multimodal.