CAT: Confidence-Adaptive Thinking for Efficient Reasoning of Large Reasoning Models

2026-07-01

AI Summary

Existing inference compression methods struggle to balance accuracy on complex tasks with efficiency on simple ones, often suffering performance degradation due to one-size-fits-all strategies. This work proposes Confidence-Adaptive Thinking (CAT), a novel framework that integrates the model’s intrinsic confidence signals into the control of reasoning chain length. By leveraging confidence-aware preference optimization, dynamic truncation, and multi-granularity difficulty-adaptive mechanisms, CAT enables differentiated processing of easy and hard questions. Experimental results demonstrate that CAT consistently outperforms current approaches across multiple benchmarks, improving reasoning accuracy on diverse base models while significantly reducing computational overhead for simpler tasks.

Abstract

Large Reasoning Models (LRMs) have achieved remarkable success on complex tasks by leveraging long chain-of-thought (CoT) trajectories, yet they frequently exhibit overthinking on simple queries, resulting in significant token overhead and reduced inference efficiency. However, existing compression methods predominantly apply uniform length reduction or rely on coarse-grained difficulty estimation, often leading to performance degradation on difficult problems. To address this limitation, we propose Confidence-Adaptive Thinking (CAT), a framework that incorporates the model's intrinsic self-certainty signals as confidence into the preference optimization process, which autonomously modulates reasoning lengths based on problem difficulty. Experimental results show that CAT consistently outperforms state-of-the-art baselines on reasoning accuracy across multiple benchmarks on different base models. Our work enables LRMs to effectively compress confident responses while deliberating on uncertain ones, offering a potentially robust solution for balancing accuracy and latency in practical industrial scenarios.

Problem

• Large Reasoning Models
• overthinking
• reasoning efficiency
• chain-of-thought compression
• difficulty-aware reasoning

Innovation

• Confidence-Adaptive Thinking
• Large Reasoning Models
• chain-of-thought compression
• preference optimization
• reasoning efficiency