1 paper · 1 filter
Lihao Zheng, Zhenwei Shao, Zhou Yu +5
Multimodal reinforcement learning with verifiable rewards commonly assigns one sequence advantage to an entire response. This outcome-level signal distinguishes overall response qu…