MM-HalBench

MM-HalBench is a multimodal hallucination benchmark used to evaluate the effectiveness of methods in reducing hallucination rates in large vision-language models.

Papers using MM-HalBench (9)

SAVE: Sparse Autoencoder-driven Visual Information Enhancement For Mitigating Object Hallucination2025Vision-Language Introspection: Mitigating Overconfident Hallucinations in MLLMs via Interpretable Bi-Causal Steering2026Context-Aware Decoding for Faithful Vision-Language Generation2026Self-Consistency as a Free Lunch: Reducing Hallucinations in Vision-Language Models via Self-Reflection2025Mitigating Hallucinations in Large Vision-Language Models via Entity-Centric Multimodal Preference Optimization2025ASCD: Attention-Steerable Contrastive Decoding for Reducing Hallucination in MLLM2025Toward More Reliable Artificial Intelligence: Reducing Hallucinations In Vision-language Models2025Hallucination Augmented Contrastive Learning for Multimodal Large Language Model2023 · 6 citesSilkie: Preference Distillation for Large Visual Language Models2023 · 3 cites