Let LLMs Break Free from Overthinking via Self-Braking Tuning

2025-05-20

AI Summary

Large reasoning models (LRMs) suffer from redundant inference and high computational overhead due to excessively long chains of thought; existing mitigation strategies largely rely on external interventions to curb “overthinking.” This paper proposes Self-Braking Tuning (SBT), the first framework enabling endogenous, external-control-free regulation of reasoning length. SBT introduces a ground-truth–based quantitative metric for overthinking, designs a self-supervised scheme for reasoning trajectory annotation and dynamic length-aware data construction, and incorporates a brake-prompt fine-tuning mechanism that empowers models to autonomously identify and terminate unproductive reasoning steps. Evaluated on mathematical reasoning benchmarks—including AIME, AMC, MATH500, and GSM8K—SBT reduces inference token consumption by up to 60% while maintaining accuracy comparable to unconstrained baselines.

Technology Category

• Knowledge Representation and Reasoning: Computational Complexity of Reasoning
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Large reasoning models (LRMs), such as OpenAI o1 and DeepSeek-R1, have significantly enhanced their reasoning capabilities by generating longer chains of thought, demonstrating outstanding performance across a variety of tasks. However, this performance gain comes at the cost of a substantial increase in redundant reasoning during the generation process, leading to high computational overhead and exacerbating the issue of overthinking. Although numerous existing approaches aim to address the problem of overthinking, they often rely on external interventions. In this paper, we propose a novel framework, Self-Braking Tuning (SBT), which tackles overthinking from the perspective of allowing the model to regulate its own reasoning process, thus eliminating the reliance on external control mechanisms. We construct a set of overthinking identification metrics based on standard answers and design a systematic method to detect redundant reasoning. This method accurately identifies unnecessary steps within the reasoning trajectory and generates training signals for learning self-regulation behaviors. Building on this foundation, we develop a complete strategy for constructing data with adaptive reasoning lengths and introduce an innovative braking prompt mechanism that enables the model to naturally learn when to terminate reasoning at an appropriate point. Experiments across mathematical benchmarks (AIME, AMC, MATH500, GSM8K) demonstrate that our method reduces token consumption by up to 60% while maintaining comparable accuracy to unconstrained models.

Problem

• Reducing redundant reasoning in large models to cut computational costs
• Enabling self-regulation in models to avoid overthinking without external control
• Maintaining accuracy while significantly decreasing token usage in reasoning tasks

Innovation

• Self-Braking Tuning for autonomous reasoning control
• Overthinking metrics based on standard answers
• Braking prompt for adaptive reasoning termination