VeriTrail: Closed-Domain Hallucination Detection with Traceability

Researchers have developed VeriTrail, the first closed-domain hallucination detection method that can trace where AI-generated misinformation originates in multi-step processes. The system addresses a critical problem where language models generate unsubstantiated content even when instructed to stick to source material, with the risk being higher in complex multi-step generative processes.

• →VeriTrail is the first hallucination detection method designed to provide traceability for both single and multi-step generative processes.
• →Multi-step generative processes have higher risks of hallucination compared to single-step processes.
• →The system can trace where hallucinated content was introduced and how faithful content was derived from source material.
• →Researchers introduced the first datasets including all intermediate outputs with human annotations for multi-step processes.
• →VeriTrail demonstrated superior performance compared to baseline hallucination detection methods.