Discover effective methods to detect hallucinations in Speech Large Language Models at inference time using attention-derived metrics for improved reliabil...
Discover HalluAudio, the first large-scale benchmark to detect hallucinations in large audio-language models across speech, music, and environmental sounds...
Discover the LLM-as-Judge framework and Ghost-100 benchmark for evaluating tone-induced hallucination in vision-language models with improved accuracy.
Discover TPA, a novel method for detecting hallucinations in Retrieval-Augmented Generation by attributing next token probabilities to key model components...