作者位次 #2/10ACL 2026Poster· Natural Language Processing
Vision-Language Introspection: Mitigating Overconfident Hallucinations in MLLM s via Interpretable Bi-Causal Steering
Shuliang Liu · Songbo Yang · Dong Fang · Sihang Jia · Yuqi Tang · Lingfeng Su · Ruoshui Peng · Yibo Yan · Xin Zou · Xuming Hu
摘要
Object hallucination critically undermines the reliability of Multimodal Large Language Models (MLLMs), often stemming from a fundamental failure in cognitive introspection—where models blindly trust linguistic priors over specific visual evidence. Existing mitigations remain limited: contrastive...