mlsanityData to 14 Oct 2026

Think or Not? Exploring Thinking Efficiency in Large Reasoning Models via an Information-Theoretic Lens

arXiv 2505.18237 ↗First posted 23 May 2025cs.CL cs.AI cs.ITcited by 34

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

Abstract (arXiv)

The recent rise of Large Reasoning Models (LRMs) has significantly improved multi-step reasoning performance, but often at the cost of generating excessively long reasoning chains. This paper revisits the efficiency of such reasoning processes through an information-theoretic lens, revealing a fundamental trade-off between reasoning length and semantic efficiency. We propose two metrics, InfoBias and InfoGain, to quantify divergence from ideal reasoning paths and stepwise information contribution, respectively. Empirical analyses show that longer reasoning chains tend to exhibit higher information bias and diminishing information gain, especially for incorrect answers. Motivated by these findings, we introduce an entropy-based Adaptive Think strategy that dynamically halts reasoning once confidence is sufficiently high, improving efficiency while maintaining competitive accuracy. Compared to the Vanilla Think approach (default mode), our strategy yields a 1.10% improvement in average accuracy and a 50.80% reduction in token usage on QwQ-32B across six benchmark tasks spanning diverse reasoning types and difficulty levels, demonstrating superior efficiency and reasoning performance. These results underscore the promise of entropy-based methods for enhancing both accuracy and cost-effiiciency in large language model deployment.

Key facts. Cited by 34 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 43 of its 64 references that are in the map. First posted on arXiv 23 May 2025.

Builds on

43 of 64 references.

Closest ancestors · 16

Same direction, newest first

  1. 2025Between Underthinking and Overthinking: An Empirical Study of Reasoning Length and correctness in LLMs30 Apr
  2. 2025Dynamic Early Exit in Reasoning Models22 Apr
  3. 2025Reasoning Models Can Be Effective Without Thinking14 Apr
  4. 2025Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models20 Mar
  5. 2025DAST: Difficulty-Adaptive Slow-Thinking for Large Reasoning Models6 Mar
  6. 2025Towards Thinking-Optimal Scaling of Test-Time Compute for LLM Reasoning25 Feb
  7. 2025The relationship between reasoning and performance in large language models--o3 (mini) thinks harder, not longer21 Feb
  8. 2025CoT-Valve: Length-Compressible Chain-of-Thought Tuning13 Feb
  9. 2025s1: Simple test-time scaling31 Jan
  10. 2025Rethinking External Slow-Thinking: From Snowball Errors to Probability of Correct Reasoning26 Jan
  11. 2024Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs30 Dec
  12. 2024Token-Budget-Aware LLM Reasoning24 Dec
4 more
  1. 2024Learning How Hard to Think: Input-Adaptive Allocation of LM Computation7 Oct
  2. 2024Make Every Penny Count: Difficulty-Adaptive Self-Consistency for Cost-Efficient Reasoning24 Aug
  3. 2024DynaThink: Fast or Slow? A Dynamic Decision-Making Framework for Large Language Models1 Jul
  4. 2024ReFT: Reasoning with Reinforced Fine-Tuning17 Jan

Other references · 27

  1. 2020Language Models are Few-Shot Learnerscited by 32,419
  2. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  3. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  4. 2020Measuring Massive Multitask Language Understandingcited by 7,963
  5. 2020Scaling Laws for Neural Language Modelscited by 6,586
  6. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,944
  7. 2024MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmarkcited by 1,848
  8. 2018CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledgecited by 1,822
  9. 2023Reasoning with Language Model is Planning with World Modelcited by 861
  10. 2022Language Models Are Greedy Reasoners: A Systematic Formal Analysis of Chain-of-Thoughtcited by 337
  11. 2025From System 1 to System 2: A Survey of Reasoning Large Language Modelscited by 320
  12. 2017Information-theoretic analysis of generalization capability of learning algorithmscited by 303
15 more
  1. 2024To CoT or not to CoT? Chain-of-thought helps mainly on math and symbolic reasoningcited by 263
  2. 2024Semantic Entropy Probes: Robust and Cheap Hallucination Detection in LLMscited by 259
  3. 2023MuSR: Testing the Limits of Chain-of-thought with Multistep Soft Reasoningcited by 201
  4. 2017Neural-Symbolic Learning and Reasoning: A Survey and Interpretationcited by 196
  5. 2025When More is Less: Understanding Chain-of-Thought Length in LLMscited by 183
  6. 2021Noisy Channel Language Model Prompting for Few-Shot Text Classificationcited by 165
  7. 2024The Impact of Reasoning Step Length on Large Language Modelscited by 153
  8. 2025Optimizing Test-Time Compute via Meta Reinforcement Fine-Tuningcited by 123
  9. 2024Prover-Verifier Games improve legibility of LLM outputscited by 85
  10. 2019BottleSum: Unsupervised and Self-supervised Sentence Summarization using the Information Bottleneck Principlecited by 39
  11. 2024Understanding Chain-of-Thought in LLMs through Information Theorycited by 37
  12. 2025BOLT: Bootstrap Long Chain-of-Thought in Language Models without Distillationcited by 23
  13. 2024CausalBN-Bench: A Comprehensive Benchmark for Causal Learning Capability of LLMscited by 20
  14. 2021Contrastive Reinforcement Learning of Symbolic Reasoning Domainscited by 17
  15. 2013Understanding ACT-R - an Outsider's Perspectivecited by 4

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 11; 2026: 23.

34 papers.

Sort:

By area:

  1. 2025Don't Overthink It: A Survey of Efficient R1-style Large Reasoning Modelscited by 37
  2. 2025Entropy After </Think> for reasoning model early exitingcited by 24
  3. 2025Towards Concise and Adaptive Thinking in Large Reasoning Models: A Surveycited by 18
  4. 2025ARES: Multimodal Adaptive Reasoning via Difficulty-Aware Token-Level Entropy Shapingcited by 17
  5. 2025DRPO: Efficient Reasoning via Decoupled Reward Policy Optimizationcited by 16
  6. 2026Parallel-Probe: Towards Efficient Parallel Thinking via 2D Probingcited by 14
  7. 2026LLMs Improving LLMs: Agentic Discovery for Test-Time Scalingcited by 10
  8. 2026Understanding Reasoning in LLMs through Strategic Information Allocation under Uncertaintycited by 9
  9. 2026FoE: Forest of Errors Makes the First Solution the Best in Large Reasoning Modelscited by 8
  10. 2025On the Provable Performance Guarantee of Efficient Reasoning Modelscited by 6
  11. 2025LYNX: Learning Dynamic Exits for Confidence-Controlled Reasoningcited by 6
  12. 2026Statistical Early Stopping for Reasoning Modelscited by 6
22 more
  1. 2026Mitigating Overthinking in Large Reasoning Language Models via Reasoning Path Deviation Monitoringcited by 6
  2. 2026Early Stopping for Large Reasoning Models via Confidence Dynamicscited by 6
  3. 2025Counterfactual Reasoning for Steerable Pluralistic Value Alignment of Large Language Modelscited by 5
  4. 2026Rethinking the Role of Entropy in Optimizing Tool-Use Behaviors for Large Language Model Agentscited by 5
  5. 2025The Zero-Step Thinking: An Empirical Study of Mode Selection as Harder Early Exit in Reasoning Modelscited by 4
  6. 2025When Reasoning Meets Its Lawscited by 3
  7. 2026VERA: Identifying and Leveraging Visual Evidence Retrieval Heads in Long-Context Understandingcited by 3
  8. 2026HypeMed: Enhancing Medication Recommendations with Hypergraph-Based Patient Relationshipscited by 3
  9. 2026How Far Ahead Do LLMs Plan? Uncovering the Latent Horizon in Chain-of-Thought Reasoningcited by 2
  10. 2026Dissecting Failure Dynamics in Large Language Model Reasoningcited by 2
  11. 2026Anytime Safe PAC Efficient Reasoningcited by 1
  12. 2026A State-Transition Framework for Efficient LLM Reasoningcited by 1
  13. 2026A*-Thought-V2: Efficient Latent Reasoning via Geometric Dynamics of LLMcited by 1
  14. 2025From Retrieval to Generation: Unifying External and Parametric Knowledge for Medical Question Answeringcited by 0
  15. 2026Conditional Performance Guarantee for Large Reasoning Modelscited by 0
  16. 2026Know When To Fold 'Em: Token-Efficient LLM Synthetic Data Generation via Multi-Stage In-Flight Rejectioncited by 0
  17. 2026Controllable Affective Generation via Latent Vector Steeringcited by 0
  18. 2026From Static Personal Values to Contextualized Personalization: Bayesian Personalized Value Alignment for LLMscited by 0
  19. 2026Learning to Stop without Learning to Stop: Self-Supervised Confidence Training Improves Reasoning Efficiencycited by 0
  20. 2026Interpreting Reasoning of Large Language Models via Partial Information Decompositioncited by 0
  21. 2026The Geometry of Knowledge Accessibility in Large Language Modelscited by 0
  22. 2026Learning to Decide, Not to Reason: Parameter-Efficient Decision Operators via Low-Rank Activation Steeringcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025THINK-Bench: Evaluating Thinking Efficiency and Chain-of-Thought Quality of Large Reasoning Models
  2. 2025Trade-offs in Large Reasoning Models: An Empirical Analysis of Deliberative and Adaptive Reasoning over Foundational Capabilities
  3. 2025The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity
  4. 2026EntroCut: Entropy-Guided Adaptive Truncation for Efficient Chain-of-Thought Reasoning in Small-scale Large Reasoning Models
  5. 2025Demystifying Reasoning Dynamics with Mutual Information: Thinking Tokens are Information Peaks in LLM Reasoning
  6. 2025Explore Briefly, Then Decide: Mitigating LLM Overthinking via Cumulative Entropy Regulation
  7. 2025Don't Think Longer, Think Wisely: Optimizing Thinking Dynamics for Large Reasoning Models
  8. 2025A*-Thought: Efficient Reasoning via Bidirectional Compression for Low-Resource Settings