Do MLLMs Really See It: Reinforcing Visual Attention in Multimodal LLMs

TL;DR

This paper identifies weak visual attention in multimodal large language models and introduces SAYO, a reinforcement learning-based approach that improves visual focus and reasoning accuracy.

Contribution

The paper proposes SAYO, a novel RL framework that explicitly aligns visual attention with reasoning steps, enhancing MLLMs' focus and reasoning capabilities.

Findings

SAYO improves performance across multiple benchmarks.

Enhanced visual attention leads to better reasoning accuracy.

Reinforcement learning effectively aligns visual focus with reasoning.

Abstract

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.