NVlabs/nano-rl
A full-stack framework for scaling reinforcement learning training of vision-language models to long video reasoning.
Not currently ranked — collecting fresh signals.
star history
Long-RL addresses the challenge of applying reinforcement learning to long video reasoning in vision-language models. It provides a 104K-sample dataset called LongVideo-Reason with high-quality reasoning annotations across diverse domains, combined with a two-stage training pipeline that handles the computational challenges of long-sequence sequence parallelism. The work produces the LongVILA-R1-7B model, demonstrating effective RL scaling to extended multi-modal contexts.
Frequently asked
- What is NVlabs/nano-rl?
- A full-stack framework for scaling reinforcement learning training of vision-language models to long video reasoning.
- Is nano-rl open source?
- Yes — NVlabs/nano-rl is open source, released under the Apache-2.0 license.
- What language is nano-rl written in?
- NVlabs/nano-rl is primarily written in Python.
- How popular is nano-rl?
- NVlabs/nano-rl has 730 stars on GitHub.
- Where can I find nano-rl?
- NVlabs/nano-rl is on GitHub at https://github.com/NVlabs/nano-rl.