AI RESEARCH

GEASS: Training-Free Caption Steering for Hallucination Mitigation in Vision-Language Models

arXiv CS.CV

ArXi:2605.01733v1 Announce Type: new Vision-Language Models (VLMs) excel at grounded reasoning but remain prone to object hallucination. Recent work treats self-generated captions as a uniformly positive resource, yet we find that naively embedding one can degrade rather than help--dropping Qwen2.5-VL-3B accuracy on HallusionBench by nearly 10 points. Two structural properties explain this. First, captions anchor not only the model's final answer but also its reasoning trajectory and lexical choices.