CAT: Confidence-Adaptive Thinking for Efficient Reasoning of Large Reasoning Models
For practitioners deploying large reasoning models, CAT offers a method to balance inference efficiency and accuracy by compressing confident responses and deliberating on uncertain ones.
CAT introduces a confidence-adaptive framework that uses the model's self-certainty signals to dynamically adjust reasoning lengths, reducing token overhead on simple queries while maintaining accuracy on difficult ones. It outperforms state-of-the-art baselines on reasoning accuracy across multiple benchmarks.
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.