Prompt-Based Monte Carlo Tree Search for Mitigating Hallucinations in Large Models

TL;DR

The paper addresses hallucinations and inefficiency when applying large language models to complex scientific problems. It introduces a prompt-based Monte Carlo Tree Search (MCTS) that dynamically adjusts exploration and adaptively selects simulation strategies, formalized via complexity($problem$) and $choose ext_simulation ext_policy(mcts ext_task, node)$ to guide reasoning. Evaluated on four SciEval subsets, the approach with Glm-4-flash+Improved MCTS achieves an average score of 65.6, outperforming GPT-3.5-Turbo–based baselines and demonstrating the viability of prompt-based MCTS for scientific tasks. The work offers a new direction for reducing hallucinations and improving reasoning efficiency in large models applied to scientific research.

Abstract

With the rapid development of large models in the field of artificial intelligence, how to enhance their application capabilities in handling complex problems in the field of scientific research remains a challenging problem to be solved. This study proposes an improved Monte Carlo Tree Search (MCTS) method based on prompt words. In the simulation search stage, it introduces dynamic adjustment of exploration parameters and adaptive selection strategies, which can better balance exploration and exploitation, thereby reducing the hallucination phenomenon. This paper takes the four subsets of the SciEval dataset as the test objects, and compares the Glm-4-flash+Improved MCTS method with the methods of several existing models. The results show that the Improved MCTS method performs better, providing new ideas and methods for the application of large models in the field of scientific research.

Figures (2)

• Figure 1: Schematic diagram of the full process of the Monte Carlo Tree Search (MCTS) method.
• Figure 2: Calculate the sum of the interior angles of a hexagon.