MISHAP-Bench: A Hallucination Benchmark for Large Audio-Language Models
This work introduces MISHAP-Bench, a comprehensive benchmark with 12,000 challenging open-ended question-audio pairs and a rigorous evaluation pipeline covering two hallucination categories, and proposes a groundedness judge that uses reference rubrics and judge prompts guided by human annotations.
Wen-Soi Zhi, Giulio Segalini, Jian-Jia Chen et al.
· 0 citations