AI RESEARCH
SAGE: Sink-Aware Grounded Decoding for Multimodal Hallucination Mitigation
arXiv CS.CV
•
ArXi:2603.27898v1 Announce Type: new Large vision-language models (VLMs) frequently suffer from hallucinations, generating content that is inconsistent with visual inputs. Existing methods typically address this problem through post-hoc filtering, additional