Entropy-Gated Dynamic Rectification for Object Hallucination in Large Vision-Language Models

Riasad Alvi, Md. Asaduzzaman Shuvo, Md. Ahashan Habib +1

Large Vision-Language Models (LVLMs) routinely name objects that are absent from the image. Latent steering can reduce this failure in one model forward pass, but a fixed gate trea…