Mitigating Action-Relation Hallucinations in LVLMs via Relation-aware Visual Enhancement
Published in ACL, 2026
Recommended citation: Qin, Z., Li, Q., Wang, Q., Qin, R., Wei, Z., & Shen, W. Mitigating Action-Relation Hallucinations in LVLMs via Relation-aware Visual Enhancement. In ACL 2026. https://arxiv.org/pdf/2605.11808
Abstract. This paper studies action-relation hallucinations in large vision-language models and proposes Relation-aware Visual Enhancement to improve attention to action-relevant image regions.
Authors: Zhenxin Qin, Qiang Li, Qingzhuo Wang, Ruiyang Qin, Zhihua Wei, Wen Shen†.