HART: Data-Driven Hallucination Attribution and Evidence-Based Tracing for Large Language Models
This work aims to improve the reliability and interpretability of LLMs for high-stakes applications by providing a fine-grained method to attribute and trace hallucinations, which is an incremental improvement over existing methods.
This paper addresses the problem of hallucinated content in LLMs by proposing HART, a framework that formalizes hallucination tracing into four stages: span localization, mechanism attribution, evidence retrieval, and causal tracing. The authors developed a structured dataset for hallucination tracing and demonstrated that HART substantially outperforms strong retrieval baselines like BM25 and DPR.
Large language models (LLMs) have demonstrated remarkable performance in text generation and knowledge-intensive question answering. Nevertheless, they are prone to producing hallucinated content, which severely undermines their reliability in high-stakes application domains. Existing hallucination attribution approaches, based on either external knowledge retrieval or internal model mechanisms, primarily focus on semantic similarity matching or representation-level discrimination. As a result, they have difficulty establishing structured correspondences at the span level between hallucination types, underlying error generation mechanisms, and external factual evidence, thereby limiting the interpretability of hallucinated fragments and the traceability of supporting or opposing evidence. To address these limitations, we propose HART, a fine-grained hallucination attribution and evidence retrieval framework for large language models. HART formalizes hallucination tracing as a structured modeling task comprising four stages: span localization, mechanism attribution, evidence retrieval, and causal tracing. Based upon this formulation, we develop the first structured dataset tailored for hallucination tracing, in which hallucination types, error mechanisms, and sets of counterfactual evidence are jointly annotated to enable causal-level interpretability evaluation. Experimental results on the proposed dataset demonstrate that HART substantially outperforms strong retrieval baselines, including BM25 and DPR, validating the effectiveness and generalization capability of the proposed tracing paradigm for hallucination analysis and evidence alignment.