Process Supervision for Chain-of-Thought Reasoning via Monte Carlo Net Information Gain

2026-03-18

AI Summary

This work addresses the challenge of error propagation in intermediate steps of multi-step reasoning, which undermines the reliability of large language models. To mitigate this issue, the authors propose a fine-grained process supervision method grounded in information theory. Their approach introduces a novel technique that leverages information gain to automatically generate step-level labels, employing Monte Carlo sampling to efficiently estimate net information gain. By integrating these labels into a process reward model, the method enhances chain-of-thought reasoning selection. Notably, the label generation complexity is reduced from O(N log N) to O(N), substantially improving scalability. Empirical evaluations across diverse domains—including mathematics, Python programming, SQL, and scientific question answering—demonstrate consistent gains in both accuracy and efficiency of best-of-K reasoning.

Technology Category

• Knowledge Representation and Reasoning: Computational Complexity of Reasoning
• Natural Language Processing: Generation
• Reasoning under Uncertainty: Stochastic Optimization

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: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Multi-step reasoning improves the capabilities of large language models (LLMs) but increases the risk of errors propagating through intermediate steps. Process reward models (PRMs) mitigate this by scoring each step individually, enabling fine-grained supervision and improved reliability. Existing methods for training PRMs rely on costly human annotations or computationally intensive automatic labeling. We propose a novel approach to automatically generate step-level labels using Information Theory. Our method estimates how each reasoning step affects the likelihood of the correct answer, providing a signal of step quality. Importantly, it reduces computational complexity to $\mathcal{O}(N)$, improving over the previous $\mathcal{O}(N \log N)$ methods. We demonstrate that these labels enable effective chain-of-thought selection in best-of-$K$ evaluation settings across diverse reasoning benchmarks, including mathematics, Python programming, SQL, and scientific question answering. This work enables scalable and efficient supervision of LLM reasoning, particularly for tasks where error propagation is critical.

Problem

• Chain-of-Thought Reasoning
• Error Propagation
• Process Supervision
• Large Language Models
• Step-level Supervision

Innovation

• Process Supervision
• Chain-of-Thought Reasoning
• Information Theory
• Monte Carlo Net Information Gain
• Step-level Reward