AI RESEARCH
TPA: Next Token Probability Attribution for Detecting Hallucinations in RAG
arXiv CS.AI
•
ArXi:2512.07515v4 Announce Type: replace-cross Detecting hallucinations in Retrieval-Augmented Generation remains a challenge. Prior approaches attribute hallucinations to a binary conflict between internal knowledge d in FFNs and the retrieved context. However, this perspective is incomplete, failing to account for the impact of other components of the LLM, such as the user query, previously generated tokens, the self token, and the final LayerNorm adjustment.