Get Started
Home
Topics
Search
Library
Research questionHow can we determine whether large audio-language models exhibit human-like responses to auditory illusions?Auditory illusions expose perceptual biases, but benchmarks for large audio-language models have largely emphasized visual illusions or general audio tasks. This makes it difficult to determine whether model responses reflect human-like perception or fidelity to the acoustic signal.
AI
Audio & Speech
Audio & Speech Processing
Evaluation & Benchmarks
Sound
Latest papersRecent research connected to this question, newest first.Auditory Illusion Benchmark for Large Audio Language ModelsThe benchmark covers ten auditory illusions across music, sound, and speech, with annotations for knowledge-based priors. It compares model responses with controlled human listening results; the reported evidence shows that models differ from humans overall, with more human-like responses in some cases involving linguistic or musical priors.research paper · Sep 2, 2026
Related questions
How can we reliably detect when an LLM response is unsupported by its reference documents?How can multimodal models rely on images or audio rather than language shortcuts?How can we detect object hallucinations in vision-language models when visual grounding shifts across layers?How can bias evaluations of large language models diagnose affected groups and reasons behind biased outputs?