SVHalluc benchmarks speech-vision hallucination in AV-LLMs

Arxiv CS.CL (NLP/language models)

SVHalluc evaluates audio-visual LLMs on speech-vision hallucination, covering object existence, attribute, and action mismatches. It complements existing benchmarks that focus only on environmental sounds.

llmmultimodalbenchmarks

See 4 more sources

Evidence Graph Consistency in Retrieval-Augmented Generation: A Model-Dependent Analysis of Hallucination Detection3 days agoJianru Shen

Hallucinations as Orthogonal Noise: Inference-Time Manifold Alignment via Dynamic Contextual Orthogonalization8 days agoMingkuan Zhao, Wentao Hu, Tianchen Huang, Yuheng Min, Suquan Chen, Yide Gao, Yanbo Zhai, Shuangyong Song, Xuelong Li

SVHalluc: Benchmarking Speech-Vision Hallucination in Audio-Visual Large Language Models8 days agoChenshuang Zhang, Kyeong Seon Kim, Chengxin Liu, Tae-Hyun Oh

Cascading Hallucination in Agentic RAG: The CHARM Framework for Detection and Mitigation7 days agoSaroj Mishra

Arxiv CS.CL (NLP/language models)·Xinyi LiZhen FangYongxin DengJinyuan LuoHongnan MaChangdae OhZijing ShiShanshan YeHanchen WangShu-Lin ChenYadan LuoMengyue YangSean DuSharon LiLing Chen

8 days ago