Can LLMs Detect Their Own Hallucinations?
This addresses the issue of hallucination detection for users relying on LLM outputs, but it is incremental as it builds on existing methods like Chain-of-Thought.
The paper tackled the problem of whether large language models (LLMs) can detect their own hallucinations, and found that GPT-3.5 Turbo with Chain-of-Thought achieved a detection rate of 58.2%.
Large language models (LLMs) can generate fluent responses, but sometimes hallucinate facts. In this paper, we investigate whether LLMs can detect their own hallucinations. We formulate hallucination detection as a classification task of a sentence. We propose a framework for estimating LLMs' capability of hallucination detection and a classification method using Chain-of-Thought (CoT) to extract knowledge from their parameters. The experimental results indicated that GPT-$3.5$ Turbo with CoT detected $58.2\%$ of its own hallucinations. We concluded that LLMs with CoT can detect hallucinations if sufficient knowledge is contained in their parameters.