Revisual-R1
Public?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.
cold-start-initializationdata-efficiencyefficient-length-rewardmathematical-reasoningmultimodal-large-language-modelopen-source-7b-modelprioritized-advantage-distillationreinforcement-learningself-reflective-chain-of-thoughtvisual-reasoning
Creat：2025-05-31T00:03:31
Update：2025-06-16T21:05:58
https://arxiv.org/abs/2506.04207
148
Stars
0
Stars Increase