Learning to Reason via Mixture-of-Thought for Logical Reasoning

TL;DR

The paper introduces Mixture-of-Thought, a multi-modality reasoning framework for LLMs that combines natural language, code, and symbolic truth-tables, significantly improving logical reasoning accuracy.

Contribution

It proposes a novel multi-modality training and inference framework, enabling LLMs to reason across three modalities, which enhances logical reasoning performance over single-modality methods.

Findings

Achieves up to +11.7pp accuracy gain on reasoning benchmarks.

Effectively handles harder logical reasoning problems.

Different modalities provide complementary reasoning strengths.

Abstract

Human beings naturally utilize multiple reasoning modalities to learn and solve logical problems, i.e., different representational formats such as natural language, code, and symbolic logic. In contrast, most existing LLM-based approaches operate with a single reasoning modality during training, typically natural language. Although some methods explored modality selection or augmentation at inference time, the training process remains modality-blind, limiting synergy among modalities. To fill in this gap, we propose Mixture-of-Thought (MoT), a framework that enables LLMs to reason across three complementary modalities: natural language, code, and a newly introduced symbolic modality, truth-table, which systematically enumerates logical cases and partially mitigates key failure modes in natural language reasoning. MoT adopts a two-phase design: (1) self-evolving MoT training, which jointly learns from filtered, self-generated rationales across modalities; and (2) MoT inference, which fully leverages the synergy of three modalities to produce better predictions. Experiments on logical reasoning benchmarks including FOLIO and ProofWriter demonstrate that our MoT framework consistently and significantly outperforms strong LLM baselines with single-modality chain-of-thought approaches, achieving up to +11.7pp average accuracy gain. Further analyses show that our MoT framework benefits both training and inference stages; that it is particularly effective on harder logical reasoning problems; and that different modalities contribute complementary strengths, with truth-table reasoning helping to overcome key bottlenecks in natural language inference.