# Reasoning in Action: MCTS-Driven Knowledge Retrieval for Large Language Models

Source: [Reasoning in Action: MCTS-Driven Knowledge Retrieval for Large Language Models](<https://www.sota2.com/research/paper/695b457853fdd5c361743005>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-01-06T05:00:04.579Z

Authors: Shuqi Liu, Bowei He, Chen Ma, Linqi Song

Published: 2025-11-28T16:23:54.000Z

## Abstract

Large language models (LLMs) typically enhance their performance through either the retrieval of semantically similar information or the improvement of their reasoning capabilities. However, a significant challenge remains in effectively integrating both retrieval and reasoning strategies to optimize LLM performance. In this paper, we introduce a reasoning-aware knowledge retrieval method that enriches LLMs with information aligned to the logical structure of conversations, moving beyond surface-level semantic similarity. We follow a coarse-to-fine approach for knowledge retrieval. First, we identify a contextually relevant sub-region of the knowledge base, ensuring that all sentences within it are relevant to the context topic. Next, we refine our search within this sub-region to extract knowledge that is specifically relevant to the reasoning process. Throughout both phases, we employ the Monte Carlo Tree Search-inspired search method to effectively navigate through knowledge sentences using common keywords. Experiments on two multi-turn dialogue datasets demonstrate that our knowledge retrieval approach not only aligns more closely with the underlying reasoning in human conversations but also significantly enhances the diversity of the retrieved knowledge, resulting in more informative and creative responses.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Knowledge Retrieval on DailyDialog](<https://www.sota2.com/research/sota/knowledge-retrieval-on-dailydialog.md>): Knowledge retrieval; DailyDialog; BERTScore Precision (avg): 84.29.

- [Knowledge Retrieval on Empathetic Dialogue (EMP)](<https://www.sota2.com/research/sota/knowledge-retrieval-on-empathetic-dialogue-emp.md>): Knowledge retrieval; Empathetic Dialogue (EMP); BERTScore Precision (avg): 85.27.

- [Human Logic Alignment on DailyDialog](<https://www.sota2.com/research/sota/human-logic-alignment-on-dailydialog.md>): Human Logic Alignment; DailyDialog; Human Logic Alignment (T=0.5): 80.97.

- [Human Logic Alignment on Empathetic Dialogue](<https://www.sota2.com/research/sota/human-logic-alignment-on-empathetic-dialogue.md>): Human Logic Alignment; Empathetic Dialogue; HLA (T=0.5): 85.2.

- [Response Generation on DailyDialog](<https://www.sota2.com/research/sota/response-generation-on-dailydialog.md>): Response Generation; DailyDialog; Pairwise Diversity: 78.5.

- [Response Generation on Empathetic Dialogue](<https://www.sota2.com/research/sota/response-generation-on-empathetic-dialogue.md>): Response Generation; Empathetic Dialogue; Pairwise Diversity: 75.

- [Dialogue Response Generation on DailyDialog](<https://www.sota2.com/research/sota/dialogue-response-generation-on-dailydialog.md>): Dialogue Response Generation; DailyDialog; Coherence: 4.08.

- [Dialogue Response Generation on Empathetic Dialogue](<https://www.sota2.com/research/sota/dialogue-response-generation-on-empathetic-dialogue.md>): Dialogue Response Generation; Empathetic Dialogue; Coherence: 3.98.

- [Knowledge Pair-wise Diversity on DailyDialog (test)](<https://www.sota2.com/research/sota/knowledge-pair-wise-diversity-on-dailydialog-test.md>): Knowledge Pair-wise Diversity; DailyDialog (test); Precision: 86.37.

- [Knowledge Pair-wise Diversity on Empathetic Dialogue (test)](<https://www.sota2.com/research/sota/knowledge-pair-wise-diversity-on-empathetic-dialogue-test.md>): Knowledge Pair-wise Diversity; Empathetic Dialogue (test); Precision: 86.53.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2601.00003>)

[PDF](<https://arxiv.org/pdf/2601.00003.pdf>)
