Stepwise Reasoning Checkpoint Analysis: A Test Time Scaling Method to Enhance LLMs' Reasoning

2025-05-23

AI Summary

Existing large language models (LLMs) rely on chain-of-thought (CoT) prompting and test-time scaling (TTS) methods—such as beam search and diverse-verifiable tree search (DVTS)—for mathematical reasoning, yet suffer from severe path homogenization and inefficient utilization of intermediate reasoning steps. To address this, we propose a stepwise reasoning checkpoint analysis framework: it dynamically inserts verifiable checkpoints between CoT steps, integrates answer-clustering-guided diverse search with checkpoint candidate enhancement to significantly improve path diversity and fault tolerance, and further aggregates multiple intermediate results via consensus-based decision making. Evaluated on standard benchmarks including GSM8K and MATH, our method consistently outperforms state-of-the-art TTS approaches, achieving substantial accuracy gains. This work establishes a novel, efficient, and robust test-time optimization paradigm for LLM-based mathematical reasoning.

Technology Category

• Search and Optimization: Learning to Search
• Machine Learning: Large Multimodal Models (LMMs)
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning

Application Category

• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Mathematical reasoning through Chain-of-Thought (CoT) has emerged as a powerful capability of Large Language Models (LLMs), which can be further enhanced through Test-Time Scaling (TTS) methods like Beam Search and DVTS. However, these methods, despite improving accuracy by allocating more computational resources during inference, often suffer from path homogenization and inefficient use of intermediate results. To address these limitations, we propose Stepwise Reasoning Checkpoint Analysis (SRCA), a framework that introduces checkpoints between reasoning steps. It incorporates two key strategies: (1) Answer-Clustered Search, which groups reasoning paths by their intermediate checkpoint answers to maintain diversity while ensuring quality, and (2) Checkpoint Candidate Augmentation, which leverages all intermediate answers for final decision-making. Our approach effectively reduces path homogenization and creates a fault-tolerant mechanism by utilizing high-quality intermediate results. Experimental results show that SRCA improves reasoning accuracy compared to existing TTS methods across various mathematical datasets.

Problem

• Reduces path homogenization in LLM reasoning steps
• Improves efficiency of intermediate result utilization
• Enhances reasoning accuracy with checkpoint strategies

Innovation

• Introduces checkpoints between reasoning steps
• Uses Answer-Clustered Search for diversity
• Leverages Checkpoint Candidate Augmentation for decisions