From Implicit to Explicit: Token-Efficient Logical Supervision for Mathematical Reasoning in LLMs

TL;DR

This paper introduces FSLR, a lightweight training framework that explicitly teaches logical relationship understanding to improve mathematical reasoning in LLMs, outperforming prior methods and reducing training costs.

Contribution

FSLR is a novel approach that explicitly supervises the first planning step in logical reasoning, addressing a key limitation of existing fine-tuning methods.

Findings

FSLR outperforms CoT-SFT with 3.2% and 4.6% accuracy improvements.

FSLR reduces training token consumption by over 80%.

FSLR achieves 4-6x faster training.

Abstract

Recent studies reveal that large language models (LLMs) exhibit limited logical reasoning abilities in mathematical problem-solving, instead often relying on pattern-matching and memorization. We systematically analyze this limitation, focusing on logical relationship understanding, which is a core capability underlying genuine logical reasoning, and reveal that errors related to this capability account for over 90\% of incorrect predictions, with Chain-of-Thought Supervised Fine-Tuning (CoT-SFT) failing to substantially reduce these errors. To address this bottleneck, we propose First-Step Logical Reasoning (FSLR), a lightweight training framework targeting logical relationship understanding. Our key insight is that the first planning step-identifying which variables to use and which operation to apply-encourages the model to derive logical relationships directly from the problem statement. By training models on this isolated step, FSLR provides explicit supervision for logical relationship understanding, unlike CoT-SFT which implicitly embeds such relationships within complete solution trajectories. Extensive experiments across multiple models and datasets demonstrate that FSLR consistently outperforms CoT-SFT under both in-distribution and out-of-distribution settings, with average improvements of 3.2\% and 4.6\%, respectively. Moreover, FSLR achieves 4-6x faster training and reduces training token consumption by over 80\%.