Structured Reasoning for Large Language Models

TL;DR

This paper introduces Structured Reasoning (SCR), a framework that enhances large language models by making reasoning processes explicit and trainable, leading to more efficient and accurate reasoning with less redundant output.

Contribution

The paper proposes SCR, a novel framework with a Generate-Verify-Revise paradigm and dynamic supervision, improving reasoning efficiency and reducing output length in large language models.

Findings

SCR improves reasoning efficiency significantly.

Reduces output token length by up to 50%.

Enhances self-verification capabilities of models.

Abstract

Large language models (LLMs) achieve strong performance by generating long chains of thought, but longer traces always introduce redundant or ineffective reasoning steps. One typical behavior is that they often perform unnecessary verification and revisions even if they have reached the correct answers. This limitation stems from the unstructured nature of reasoning trajectories and the lack of targeted supervision for critical reasoning abilities. To address this, we propose Structured Reasoning (SCR), a framework that decouples reasoning trajectories into explicit, evaluable, and trainable components. We mainly implement SCR using a Generate-Verify-Revise paradigm. Specifically, we construct structured training data and apply Dynamic Termination Supervision to guide the model in deciding when to terminate reasoning. To avoid interference between learning signals for different reasoning abilities, we adopt a progressive two-stage reinforcement learning strategy: the first stage targets initial generation and self-verification, and the second stage focuses on revision. Extensive experiments on three backbone models show that SCR substantially improves reasoning efficiency and self-verification. Besides, compared with existing reasoning paradigms, it reduces output token length by up to 50%.