Every Token Leaves a Ripple in the Stream of Thought: Eliciting Model-Internal Token Saliency for Chain-of-Thought Compression

• URL: http://arxiv.org/abs/2608.31066v1
• Date: Mon, 31 Aug 2026 16:41:14 GMT
• Title: Every Token Leaves a Ripple in the Stream of Thought: Eliciting Model-Internal Token Saliency for Chain-of-Thought Compression
• Abstract summary: Chain-of-thought (CoT) reasoning improves multi-step problem solving, but long reasoning traces inflate computation cost.<n>We propose textscMIST (Model-Internal Saliency for Token-level CoT compression), which defines token importance along two complementary axes.<n>textscMIST consistently outperforms baseline methods, suggesting that model-internal saliency provides an effective proxy for reasoning-token importance.
• Score: 13.74199057436712
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Chain-of-thought (CoT) reasoning improves multi-step problem solving, but long reasoning traces inflate inference cost. Token-level CoT compression reduces this cost by pruning full reasoning chains into shorter traces for model adaptation, making token selection the central challenge. Existing methods often rely on external scorers or heuristic signals only indirectly tied to the model's internal answer computation. We instead adopt a model-internal perspective: as the model forms an answer, each reasoning token leaves a ripple in the residual stream, the model's \emph{stream of thought}, and the magnitude of this ripple reflects the token's contribution to the answer computation. Building on this view, we propose \textsc{MIST} (Model-Internal Saliency for Token-level CoT compression), which defines token importance along two complementary axes: \emph{necessity}, the drop in answer likelihood when a token's internal contribution is removed, and \emph{sufficiency}, the gain in answer likelihood when that contribution alone is provided. Combining the two yields a unified importance score for pruning. Across four reasoning benchmarks and four models, \textsc{MIST} consistently outperforms baseline methods, suggesting that model-internal saliency provides an effective proxy for reasoning-token importance.

Related papers

• C$^{2}$R: Cross-sample Consistency Regularization Mitigates Feature Splitting and Absorption in Sparse Autoencoders [56.67585330425431]
  Sparse Autoencoders (SAEs) are widely used to interpret large language models by decomposing activations into sparse, human-understandable features.<n>These issues stem from inconsistent latent assignment across samples.<n>We introduce C$2$R, which penalizes the co-activation of directionally similar latents.
  arXiv  Detail & Related papers  (2026-06-29T17:45:31Z)
• Shorthand for Thought: Compressing LLM Reasoning via Entropy-Guided Supertokens [4.0704009036918025]
  Reasoning in Large Language Models incurs significant inference-time compute.<n>We show that reasoning tokens split into two functional types: low-entropy textitstructural tokens (recurring phrases that scaffold the reasoning process) and higher-entropy textitorganic tokens (problem-specific content that drives toward a solution)<n>We derive textitsupertokens that capture frequent structural patterns, then teach the model to adopt them via supervised fine-tuning.
  arXiv  Detail & Related papers  (2026-04-29T07:06:43Z)
• CRISP: Compressing Redundancy in Chain-of-Thought via Intrinsic Saliency Pruning [25.840307083323456]
  Long Chain-of-Thought (CoT) reasoning is pivotal for the success of recent reasoning models but suffers from high computational overhead and latency.<n>This paper presents textbfCompressing textbfRedundancy in Chain-of-Thought via textbfIntrinsic textbfSaliency textbfPruning (textbfCRISP), a framework that compresses CoT by exploiting the model's intrinsic saliency.
  arXiv  Detail & Related papers  (2026-04-19T07:26:55Z)
• Accordion-Thinking: Self-Regulated Step Summaries for Efficient and Readable LLM Reasoning [62.680551162054975]
  We introduce an end-to-end framework where LLMs learn to self-regulate the granularity of the reasoning steps through dynamic summarization.<n>We apply reinforcement learning to incentivize this capability further, uncovering a critical insight: the accuracy gap between the highly efficient Fold mode and the exhaustive Unfold mode progressively narrows.<n>Our Accordion-Thinker demonstrates that with learned self-compression, LLMs can tackle complex reasoning tasks with minimal dependency token overhead.
  arXiv  Detail & Related papers  (2026-02-03T08:34:20Z)
• P2S: Probabilistic Process Supervision for General-Domain Reasoning Question Answering [51.04492568024515]
  We introduce Probabilistic Process Supervision (P2S), a novel framework for fine-grained process rewards.<n>P2S provides fine-grained process rewards without requiring a separate reward model or human-annotated reasoning steps.
  arXiv  Detail & Related papers  (2026-01-28T14:35:20Z)
• Do LLMs Encode Functional Importance of Reasoning Tokens? [11.21558453188654]
  We propose greedy pruning, a likelihood-preserving deletion procedure that iteratively removes reasoning tokens.<n>We show that students trained on pruned chains outperform a frontier-model-supervised compression baseline at matched reasoning lengths.
  arXiv  Detail & Related papers  (2026-01-06T14:50:02Z)
• In-Token Rationality Optimization: Towards Accurate and Concise LLM Reasoning via Self-Feedback [38.915062716409686]
  InTRO is a new framework that enables both token-level exploration and self-feedback for accurate and concise reasoning.<n>InTRO consistently outperforms other baselines, raising solution accuracy by up to 20% relative to the base model.<n>Its chains of thought are notably more concise, exhibiting reduced verbosity.
  arXiv  Detail & Related papers  (2025-11-13T01:47:06Z)
• ConciseHint: Boosting Efficient Reasoning via Continuous Concise Hints during Generation [74.37307916314407]
  We propose a framework dubbed ConciseHint, which continuously encourages the reasoning model to speak concisely.<n>Experiments on the state-of-the-art LRMs, including DeepSeek-R1 and Qwen-3 series, demonstrate that our method can effectively produce concise reasoning.
  arXiv  Detail & Related papers  (2025-06-23T16:20:44Z)
• Fractured Chain-of-Thought Reasoning [61.647243580650446]
  We introduce Fractured Sampling, a unified inference-time strategy that interpolates between full CoT and solution-only sampling.<n>We show that Fractured Sampling consistently achieves superior accuracy-cost trade-offs, yielding steep log-linear scaling gains in Pass@k versus token budget.
  arXiv  Detail & Related papers  (2025-05-19T11:30:41Z)
• ConCISE: Confidence-guided Compression in Step-by-step Efficient Reasoning [64.93140713419561]
  Large Reasoning Models (LRMs) perform strongly in complex reasoning tasks via Chain-of-Thought (CoT) prompting, but often suffer from verbose outputs.<n>Existing fine-tuning-based compression methods either operate post-hoc pruning, risking disruption to reasoning coherence, or rely on sampling-based selection.<n>We introduce ConCISE, a framework designed to generate concise reasoning chains, integrating Confidence Injection to boost reasoning confidence, and Early Stopping to terminate reasoning when confidence is sufficient.
  arXiv  Detail & Related papers  (2025-05-08T01:40:40Z)
• CoT-Valve: Length-Compressible Chain-of-Thought Tuning [50.196317781229496]
  We introduce a new tuning and inference strategy named CoT-Valve, designed to allow models to generate reasoning chains of varying lengths.<n>We show that CoT-Valve successfully enables controllability and compressibility of the chain and shows better performance than the prompt-based control.
  arXiv  Detail & Related papers  (2025-02-13T18:52:36Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.