mlsanityData to 14 Oct 2026

From Long to Lean: Performance-aware and Adaptive Chain-of-Thought Compression via Multi-round Refinement

arXiv 2509.22144 ↗First posted 26 Sep 2025cs.CL cs.AIcited by 2

On the map: LLM reasoning → Test-time compute and efficient thinking

Abstract (arXiv)

Chain-of-Thought (CoT) reasoning improves performance on complex tasks but introduces significant inference latency due to verbosity. We propose Multiround Adaptive Chain-of-Thought Compression (MACC), a framework that leverages the token elasticity phenomenon--where overly small token budgets can paradoxically increase output length--to progressively compress CoTs via multiround refinement. This adaptive strategy allows MACC to determine the optimal compression depth for each input. Our method achieves an average accuracy improvement of 5.6 percent over state-of-the-art baselines, while also reducing CoT length by an average of 47 tokens and significantly lowering latency. Furthermore, we show that test-time performance--accuracy and token length--can be reliably predicted using interpretable features like perplexity and compression rate on the training set. Evaluated across different models, our method enables efficient model selection and forecasting without repeated fine-tuning, demonstrating that CoT compression is both effective and predictable. Our code will be released in https://github.com/Leon221220/MACC.

Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 45 of its 50 references that are in the map. First posted on arXiv 26 Sep 2025.

Builds on

45 of 50 references.

Closest ancestors · 19

Same direction, newest first

  1. 2025Efficient Inference for Large Reasoning Models: A Survey29 Mar
  2. 2025A Survey of Efficient Reasoning for Large Reasoning Models: Language, Multimodality, and Beyond27 Mar
  3. 2025Think Twice: Enhancing LLM Reasoning by Scaling Multi-round Test-time Thinking25 Mar
  4. 2025InftyThink: Breaking the Length Limits of Long-Context Reasoning in Large Language Models9 Mar
  5. 2025DAST: Difficulty-Adaptive Slow-Thinking for Large Reasoning Models6 Mar
  6. 2025Meta-Reasoner: Dynamic Guidance for Optimized Inference-time Reasoning in Large Language Models27 Feb
  7. 2025Towards Thinking-Optimal Scaling of Test-Time Compute for LLM Reasoning25 Feb
  8. 2025LightThinker: Thinking Step-by-Step Compression21 Feb
  9. 2025TokenSkip: Controllable Chain-of-Thought Compression in LLMs17 Feb
  10. 2025Demystifying Long Chain-of-Thought Reasoning in LLMs5 Feb
  11. 2025Efficient Reasoning with Hidden Thinking31 Jan
  12. 2025O1-Pruner: Length-Harmonizing Fine-Tuning for O1-Like Reasoning Pruning22 Jan
7 more
  1. 2024Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs30 Dec
  2. 2024Token-Budget-Aware LLM Reasoning24 Dec
  3. 2024C3oT: Generating Shorter Chain-of-Thought without Compromising Effectiveness16 Dec
  4. 2024Language Models are Hidden Reasoners: Unlocking Latent Reasoning Capabilities via Self-Rewarding6 Nov
  5. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters6 Aug
  6. 2024Reasoning in Token Economies: Budget-Aware Evaluation of LLM Reasoning Strategies10 Jun
  7. 2023Tree of Thoughts: Deliberate Problem Solving with Large Language Models17 May

Other references · 26

  1. 2024The Llama 3 Herd of Modelscited by 14,715
  2. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,380
  3. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  4. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  5. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,330
  6. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,708
  7. 2022Large Language Models are Zero-Shot Reasonerscited by 4,374
  8. 2023Let's Verify Step by Stepcited by 3,789
  9. 2022FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awarenesscited by 3,673
  10. 2024Qwen2 Technical Reportcited by 2,120
  11. 2022Least-to-Most Prompting Enables Complex Reasoning in Large Language Modelscited by 1,407
  12. 2024OpenAI o1 System Cardcited by 1,244
14 more
  1. 2021Show Your Work: Scratchpads for Intermediate Computation with Language Modelscited by 886
  2. 2024Training Large Language Models to Reason in a Continuous Latent Spacecited by 797
  3. 2022Large Language Models Are Reasoning Teacherscited by 329
  4. 2025Chain of Draft: Thinking Faster by Writing Lesscited by 297
  5. 2024Distilling System 2 into System 1cited by 159
  6. 2025Optimizing Test-Time Compute via Meta Reinforcement Fine-Tuningcited by 123
  7. 2025Sketch-of-Thought: Efficient LLM Reasoning with Adaptive Cognitive-Inspired Sketchingcited by 117
  8. 2024RetrievalAttention: Accelerating Long-Context LLM Inference via Vector Retrievalcited by 116
  9. 2023Response Length Perception and Sequence Scheduling: An LLM-Empowered LLM Inference Pipelinecited by 86
  10. 2024Can Language Models Learn to Skip Steps?cited by 59
  11. 2025Stepwise Perplexity-Guided Refinement for Efficient Chain-of-Thought Reasoning in Large Language Modelscited by 51
  12. 2023What's the Magic Word? A Control Theory of LLM Promptingcited by 41
  13. 2024Over-Reasoning and Redundant Calculation of Large Language Modelscited by 27
  14. 2024Verbosity-Aware Rationale Reduction: Effective Reduction of Redundant Rationale via Principled Criteriacited by 4

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026Learning from Partial Chain-of-Thought via Truncated-Reasoning Self-Distillationcited by 3
  2. 2026MOAE: Multi-Objective Agent Evolution with Pareto-Preserving Searchcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025R1-Compress: Long Chain-of-Thought Compression via Chunk Compression and Search
  2. 2025Reasoning Efficiently Through Adaptive Chain-of-Thought Compression: A Self-Optimizing Framework
  3. 2025CoT-Valve: Length-Compressible Chain-of-Thought Tuning
  4. 2026Towards Efficient Large Language Reasoning Models via Extreme-Ratio Chain-of-Thought Compression
  5. 2026HybridThinker: Efficient Chain-of-Thought Reasoning via Compressed Memory and Transient Thought Steps
  6. 2026Chain Of Thought Compression: A Theoretical Analysis
  7. 2026When Compression Helps and When It Hurts: Condition-Aware Analysis of Chain-of-Thought Distillation
  8. 2026MemoSight: Unifying Context Compression and Multi Token Prediction for Reasoning Acceleration