Large Language Models Can Self-Correct with Key Condition Verification
This addresses the limitation of LLMs in self-correcting reasoning errors, offering a simple, incremental improvement for AI systems handling complex tasks.
The paper tackles the problem of large language models (LLMs) failing to self-correct reasoning by proposing ProCo, a verification method that masks key conditions to verify and iteratively correct responses, resulting in average improvements of +6.8 exact match on QA, +14.1 accuracy on arithmetic, and +9.6 accuracy on commonsense reasoning compared to prior methods.
Intrinsic self-correct was a method that instructed large language models (LLMs) to verify and correct their responses without external feedback. Unfortunately, the study concluded that the LLMs could not self-correct reasoning yet. We find that a simple yet effective verification method can unleash inherent capabilities of the LLMs. That is to mask a key condition in the question, add the current response to construct a verification question, and predict the condition to verify the response. The condition can be an entity in an open-domain question or a numeric value in a math question, which requires minimal effort (via prompting) to identify. We propose an iterative verify-then-correct framework to progressively identify and correct (probably) false responses, named ProCo. We conduct experiments on three reasoning tasks. On average, ProCo, with GPT-3.5-Turbo as the backend LLM, yields $+6.8$ exact match on four open-domain question answering datasets, $+14.1$ accuracy on three arithmetic reasoning datasets, and $+9.6$ accuracy on a commonsense reasoning dataset, compared to Self-Correct. Our implementation is made publicly available at https://wzy6642.github.io/proco.github.io/.