Back to search
[ICLR 2026]🚀ReVisual-R1 is a 7B open-source multimodal language model that follows a three-stage curriculum—cold-start pre-training, multimodal reinforcement learning, and text-only reinforcement learning—to achieve faithful, concise, and self-reflective state-of-the-art performance in visual and textual reasoning.
Stars
206
Forks
3
Watchers
206
Open Issues
10
Overall repository health assessment
No package.json found
This might not be a Node.js project