The Illusion of Progress: Re-evaluating Hallucination Detection in LLMs
This work addresses the reliability of LLM deployment by exposing flaws in evaluation practices, which is incremental as it critiques existing methods rather than introducing new ones.
The paper tackled the problem of hallucination detection in LLMs by showing that current evaluations using ROUGE are misleading, with performance drops of up to 45.9% when using human-aligned metrics, and revealed that simple heuristics can rival complex methods.
Large language models (LLMs) have revolutionized natural language processing, yet their tendency to hallucinate poses serious challenges for reliable deployment. Despite numerous hallucination detection methods, their evaluations often rely on ROUGE, a metric based on lexical overlap that misaligns with human judgments. Through comprehensive human studies, we demonstrate that while ROUGE exhibits high recall, its extremely low precision leads to misleading performance estimates. In fact, several established detection methods show performance drops of up to 45.9\% when assessed using human-aligned metrics like LLM-as-Judge. Moreover, our analysis reveals that simple heuristics based on response length can rival complex detection techniques, exposing a fundamental flaw in current evaluation practices. We argue that adopting semantically aware and robust evaluation frameworks is essential to accurately gauge the true performance of hallucination detection methods, ultimately ensuring the trustworthiness of LLM outputs.