Critic-CoT: Boosting the reasoning abilities of large language model via Chain-of-thoughts Critic
This work addresses the challenge of improving reasoning abilities in LLMs for tasks like math problem-solving, though it appears incremental by building on existing self-critic and CoT methods.
The paper tackles the problem of limited reasoning capabilities in large language models by proposing Critic-CoT, a framework that enhances self-critique through step-wise reasoning and automatic data construction, resulting in significant performance boosts on GSM8K and MATH benchmarks.
Self-critic has become a crucial mechanism for enhancing the reasoning performance of LLMs. However, current approaches mainly involve basic prompts for intuitive instance-level feedback, which resembles System-1 processes and limits the reasoning capabilities. Moreover, there is a lack of in-depth investigations into the relationship between LLM's ability to criticize and its task-solving performance. To address these issues, we propose Critic-CoT, a novel framework that pushes LLMs toward System-2-like critic capability. Through a step-wise CoT reasoning paradigm and the automatic construction of distant-supervision data without human annotation, Critic-CoT enables LLMs to engage in slow, analytic self-critique and refinement, thereby improving their reasoning abilities. Experiments on GSM8K and MATH demonstrate that our enhanced model significantly boosts task-solving performance by filtering out invalid solutions or iterative refinement. Furthermore, we investigate the intrinsic correlation between critique and task-solving abilities within LLMs, discovering that these abilities can mutually reinforce each other rather than conflict.