MARCO: Meta-Reflection with Cross-Referencing for Code Reasoning
This addresses the challenge of static, isolated problem-solving in code reasoning for AI systems, offering a dynamic, cognitive-evolving approach.
The paper tackles the problem of improving large language models' code reasoning by enabling them to self-improve during inference, achieving substantial cumulative gains as demonstrated across various datasets.
The ability to reason is one of the most fundamental capabilities of large language models (LLMs), enabling a wide range of downstream tasks through sophisticated problem-solving. A critical aspect of this is code reasoning, which involves logical reasoning with formal languages (i.e., programming code). In this paper, we enhance this capability of LLMs by exploring the following question: how can an LLM agent become progressively smarter in code reasoning with each solution it proposes, thereby achieving substantial cumulative improvement? Most existing research takes a static perspective, focusing on isolated problem-solving using frozen LLMs. In contrast, we adopt a cognitive-evolving perspective and propose a novel framework named Meta-Reflection with Cross-Referencing (MARCO) that enables the LLM to evolve dynamically during inference through self-improvement. From the perspective of human cognitive development, we leverage both knowledge accumulation and lesson sharing. In particular, to accumulate knowledge during problem-solving, we propose meta-reflection that reflects on the reasoning paths of the current problem to obtain knowledge and experience for future consideration. Moreover, to effectively utilize the lessons from other agents, we propose cross-referencing that incorporates the solution and feedback from other agents into the current problem-solving process. We conduct experiments across various datasets in code reasoning, and the results demonstrate the effectiveness of MARCO.