How Language Model Hallucinations Can Snowball
This addresses the risk of hallucination snowballing in language models for practical applications, though it is incremental as it builds on known issues of hallucinations.
The paper tackles the problem of language models hallucinating incorrect statements by showing that they often generate false claims to justify earlier hallucinations, which they can separately recognize as incorrect, with ChatGPT and GPT-4 identifying 67% and 87% of their own mistakes, respectively.
A major risk of using language models in practical applications is their tendency to hallucinate incorrect statements. Hallucinations are often attributed to knowledge gaps in LMs, but we hypothesize that in some cases, when justifying previously generated hallucinations, LMs output false claims that they can separately recognize as incorrect. We construct three question-answering datasets where ChatGPT and GPT-4 often state an incorrect answer and offer an explanation with at least one incorrect claim. Crucially, we find that ChatGPT and GPT-4 can identify 67% and 87% of their own mistakes, respectively. We refer to this phenomenon as hallucination snowballing: an LM over-commits to early mistakes, leading to more mistakes that it otherwise would not make.