MARCO: Meta-Reflection with Cross-Referencing for Code Reasoning
Yusheng Zhao, Xiao Luo, Weizhi Zhang, Wei Ju, Zhiping Xiao, Philip S. Yu, Ming Zhang

TL;DR
This paper introduces MARCO, a novel framework that enables large language models to improve their code reasoning capabilities dynamically through self-reflection and cross-referencing, leading to better problem-solving performance.
Contribution
We propose MARCO, a framework that allows LLMs to evolve during inference by reflecting on their reasoning and leveraging solutions from other agents, which is a significant advancement over static problem-solving methods.
Findings
MARCO improves code reasoning accuracy across multiple datasets.
Self-reflection enhances the model's problem-solving capabilities.
Cross-referencing effectively incorporates external solutions for better reasoning.
Abstract
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…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
MethodsADaptive gradient method with the OPTimal convergence rate
