Do MLLMs Really See It: Reinforcing Visual Attention in Multimodal LLMs
This addresses a specific bottleneck in MLLMs for researchers and practitioners, offering an incremental improvement in visual attention mechanisms.
The paper tackles the problem of weak visual attention in multimodal large language models (MLLMs), which leads to error propagation and failed inferences, and proposes SAYO, a model trained with reinforcement learning that improves performance on diverse reasoning and perception tasks across multiple benchmarks.
While chain-of-thought (CoT) reasoning has substantially improved multimodal large language models (MLLMs) on complex reasoning tasks, existing approaches largely rely on long textual reasoning trajectories and provide limited mechanisms for learning stable visual attention policies. Our analysis shows that current MLLMs exhibit weak visual focus: early-stage visual misalignment is rarely corrected during subsequent reasoning, leading to error propagation and failed inferences. We argue that this limitation stems from inadequate credit assignment for visual attention during training. To address this issue, we propose SAYO, a visual reasoning model trained with a reinforcement learning (RL) framework that introduces a region-level visual attention-based reward. This reward explicitly aligns optimization signals with visually grounded reasoning steps, enabling the model to learn more reliable attention behaviors. Extensive experiments across multiple multimodal benchmarks demonstrate that SAYO consistently improves performance on diverse reasoning and perception tasks.