Agents-X/

PyVision-Video-RL-Data

Project Page | Paper | GitHub This repository contains the reinforcement learning (RL) data used to train PyVision-Video-RL, as presented in the paper PyVision-RL: Forging Open Agentic Vision Models via RL. PyVision-RL is a reinforcement learning framework for open-weight multimodal models that stabilizes training and sustains interaction. For video reasoning, PyVision-Video employs on-demand context construction, selectively sampling task-relevant frames… See the full description on the dataset page: https://huggingface.co/datasets/Agents-X/PyVision-Video-RL-Data.