AI RESEARCH

Detecting Hallucinations in SpeechLLMs at Inference Time Using Attention Maps

arXiv CS.AI

ArXi:2604.19565v1 Announce Type: cross Hallucinations in Speech Large Language Models (SpeechLLMs) pose significant risks, yet existing detection methods typically rely on gold-standard outputs that are costly or impractical to obtain. Moreover, hallucination detection methods developed for text-based LLMs do not directly capture audio-specific signals.