AIBullisharXiv – CS AI · 6h ago6/10
🧠
Density Ridge Selective Prediction for LLM and VLM Hallucination Detection under Calibration Label Scarcity
Researchers propose a density ridge-based method for detecting hallucinations in large language and vision-language models that outperforms existing approaches by 5-20 AUROC points while requiring minimal calibration labels. The technique maps hidden state trajectories to a low-dimensional geometric skeleton, enabling robust hallucination detection even when training data is scarce.