一作ACL 2026Poster· Natural Language Processing
Visual Attention Reasoning via Hierarchical Search and Self-Verification
Wei Cai · Jian Zhao · Yuchen Yuan · Tianle Zhang · Ming Zhu · HAICHUAN TANG · Xuelong Li
摘要
Multimodal Large Language Models (MLLMs) frequently hallucinate due to their reliance on fragile, linear reasoning and weak visual grounding. We propose Visual Attention Reasoning (VAR), a reinforcement learning framework that reformulates reasoning as a hierarchical search with self-verification...