一作ACL 2026Poster· Natural Language Processing
LAF a CT : Attribution-based Localization and Focused Sequential Analysis of Fact-Critical Tokens for Hallucination Detection
Xin Wang · Jiahao Li · Licheng Zhang · Zhendong Mao
摘要
Large Language Models (LLMs) suffer from hallucinations, severely undermining their reliability. While white-box hallucination detection methods that leverage hidden states prevail, they fail to identify and focus on fact-critical information when analyzing token sequences. To address this, we pr...