Rethinking Token-Level Policy Optimization for Multimodal Chain-of-Thought
This work addresses the challenge of fine-grained reasoning in multimodal AI for tasks requiring visual grounding and inference, representing an incremental improvement over existing RLVR frameworks.
The paper tackles the problem of optimizing multimodal Chain-of-Thought reasoning by addressing coarse granularity in existing methods, proposing Perception-Exploration Policy Optimization (PEPO) which improves performance across diverse benchmarks like geometry reasoning and visual puzzle solving, achieving consistent gains over strong baselines.
Multimodal Chain-of-Thought (CoT) reasoning requires large vision-language models to construct reasoning trajectories that interleave perceptual grounding with multi-step inference. However, existing Reinforcement Learning with Verifiable Rewards (RLVR) methods typically optimize reasoning at a coarse granularity, treating CoT uniformly without distinguishing their varying degrees of visual grounding. In this work, we conduct a token-level analysis of multimodal reasoning trajectories and show that successful reasoning is characterized by structured token dynamics reflecting both perceptual grounding and exploratory inference. Building upon this analysis, we propose Perception-Exploration Policy Optimization (PEPO), which derives a perception prior from hidden state similarity and integrates it with token entropy through a smooth gating mechanism to produce token-level advantages. PEPO integrates seamlessly with existing RLVR frameworks such as GRPO and DAPO, requiring neither additional supervision nor auxiliary branches. Extensive experiments across diverse multimodal benchmarks demonstrate consistent and robust improvements over strong RL baselines, spanning geometry reasoning, visual grounding, visual puzzle solving, and few-shot classification, while maintaining stable training dynamics. Code: https://github.com/xzxxntxdy/PEPO