CRISP: Compressing Redundancy in Chain-of-Thought via Intrinsic Saliency Pruning

arXiv:2604.17297v1 Announce Type: new Abstract: Long Chain-of-Thought (CoT) reasoning is pivotal for the success of recent reasoning models but suffers from high computational overhead and latency. While prior works attempt to compress CoT via external compressor, they often fail to align with the model's internal reasoning dynamics, resulting in the loss of critical logical steps. This paper presents extbf{C}ompressing extbf{R}edundancy in Chain-of-Thought via extbf{I}ntrinsic extbf{S}aliency extbf{P}runing (extbf{CRISP}), a framework that compresses CoT by exploiting the model's intrinsic saliency. Our analysis reveals a distinct phenomenon: the reasoning termination token exttt{[object Object]} acts as an information anchor, where its attention pattern effectively demarcates essential reasoning from redundancy. Based on this finding, we design a policy that utilizes these intrinsic attention signals to guide atomic compression operations. In contrast to coarse-grained pruning strategies, CRISP strategically distills the reasoning chain to maximize information density while preserving logical coherence. Empirical results across various backbone models and mathematical datasets demonstrate that CRISP achieves a 50-60% reduction in token count without compromising accuracy, effectively mitigating the efficiency bottleneck of long-context reasoning. We open-source our implementation to facilitate further research in efficient reasoning.

Related

• Think Less, Know More: State-Aware Reasoning Compression with Knowledge Guidance for Efficient Reasoning
• Graph-Based Chain-of-Thought Pruning for Reducing Redundant Reflections in Reasoning LLMs
• Not All Tokens Matter: Towards Efficient LLM Reasoning via Token Significance in Reinforcement Learning
• ORBIT: On-policy Exploration-Exploitation for Controllable Multi-Budget Reasoning

Source: arXiv cs.CL | 2026-04-21