Table-Critic: A Multi-Agent Framework for Collaborative Criticism and Refinement in Table Reasoning

2025-02-17

AI Summary

To address the cascading failure problem in multi-step tabular reasoning with large language models (LLMs)—caused by error propagation from erroneous intermediate steps—this paper proposes a four-role collaborative multi-agent framework (Judge-Critic-Refiner-Curator) that establishes a dynamic “critique–refine–converge” diagnostic and iterative optimization mechanism. It introduces, for the first time, an auto-evolving template tree structure to enable experience-based accumulation and cross-task transfer of critique knowledge, integrated with reflective reasoning, template-tree-driven metacognitive learning, and role-aware prompt engineering. Evaluated on multiple tabular reasoning benchmarks, the method significantly improves accuracy and error correction rates while reducing solution degeneration, under controllable computational overhead. Key contributions include: (1) an evolvable error-correction paradigm; (2) a closed-loop, role-coordinated multi-agent mechanism; and (3) a metacognitive reasoning architecture grounded in the template tree.

Technology Category

• Multiagent Systems: Multiagent Learning
• Search and Optimization: Metareasoning and Metaheuristics
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Despite the remarkable capabilities of large language models (LLMs) in various reasoning tasks, they still struggle with table reasoning tasks, particularly in maintaining consistency throughout multi-step reasoning processes. While existing approaches have explored various decomposition strategies, they often lack effective mechanisms to identify and correct errors in intermediate reasoning steps, leading to cascading error propagation. To address these issues, we propose Table-Critic, a novel multi-agent framework that facilitates collaborative criticism and iterative refinement of the reasoning process until convergence to correct solutions. Our framework consists of four specialized agents: a Judge for error identification, a Critic for comprehensive critiques, a Refiner for process improvement, and a Curator for pattern distillation. To effectively deal with diverse and unpredictable error types, we introduce a self-evolving template tree that systematically accumulates critique knowledge through experience-driven learning and guides future reflections. Extensive experiments have demonstrated that Table-Critic achieves substantial improvements over existing methods, achieving superior accuracy and error correction rates while maintaining computational efficiency and lower solution degradation rate.

Problem

• Improves table reasoning in LLMs
• Reduces error propagation in multi-step reasoning
• Enhances accuracy with multi-agent collaboration

Innovation

• Multi-agent collaborative criticism
• Self-evolving template tree
• Iterative error correction refinement