mlsanityData to 14 Oct 2026

PathCal: State-Aware Reflection-Marker Calibration for Efficient Reasoning

arXiv 2605.23074 ↗First posted 21 May 2026cs.AIcited by 0

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

Abstract (arXiv)

Large Reasoning Language Models (LRMs) tackle complex tasks through test-time scaling, generating long-form Chain-of-Thought (CoT) trajectories during inference. These trajectories contain reflection markers such as wait, but, and alternatively, accompanying hesitation, revision, and alternative explorations. Recent test-time control methods use these markers to steer reasoning, but treating them as one coarse-grained category can obscure their distinct intervention effects. We study how marker identity and reasoning context shape these effects. Type-wise suppression produces different accuracy--length tradeoffs, while fixed-prefix interventions show that the relative benefit of continuation and revision varies across prefixes and tasks. Motivated by these findings, we introduce PathCal, a training-free controller that adjusts the strength of a fixed continuation bias using instantaneous competition among continuation, revision, and alternative-opening markers. It applies finite logit adjustments without additional model forward passes. An exact log-odds identity and frozen-prefix probes establish its local directional effect. Experiments show that PathCal improves the accuracy--length tradeoff in many mathematical reasoning settings and also reduces generation length in coding and multilingual mathematics.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 58 of its 75 references that are in the map. First posted on arXiv 21 May 2026.

Builds on

58 of 75 references.

Closest ancestors · 27

Same direction, newest first

  1. 2026CAPS: Cascaded Adaptive Pairwise Selection for Efficient Parallel Reasoning15 May
  2. 2026Thinking by Subtraction: Confidence-Driven Contrastive Decoding for LLM Reasoning20 Feb
  3. 2026Directional Reasoning Trajectory Change (DRTC): Identifying Critical Trace Segments in Reasoning Models17 Feb
  4. 2025First Try Matters: Revisiting the Role of Reflection in Reasoning Models9 Oct
  5. 2025It's Not That Simple. An Analysis of Simple Test-Time Scaling19 Jul
  6. 2025Reasoning-Finetuning Repurposes Latent Representations in Base Models16 Jul
  7. 2025Do Thinking Tokens Help or Trap? Towards More Efficient Large Reasoning Model30 Jun
  8. 2025AdapThink: Adaptive Thinking Preferences for Reasoning Language Model23 Jun
  9. 2025Demystifying Reasoning Dynamics with Mutual Information: Thinking Tokens are Information Peaks in LLM Reasoning3 Jun
  10. 2025A*-Thought: Efficient Reasoning via Bidirectional Compression for Low-Resource Settings30 May
  11. 2025AdaCtrl: Towards Adaptive and Controllable Reasoning via Difficulty-Aware Budgeting24 May
  12. 2025R1-Compress: Long Chain-of-Thought Compression via Chunk Compression and Search22 May
15 more
  1. 2025ConCISE: Confidence-guided Compression in Step-by-step Efficient Reasoning8 May
  2. 2025Between Underthinking and Overthinking: An Empirical Study of Reasoning Length and correctness in LLMs30 Apr
  3. 2025Dynamic Early Exit in Reasoning Models22 Apr
  4. 2025Understanding Aha Moments: from External Observations to Internal Mechanisms3 Apr
  5. 2025L1: Controlling How Long A Reasoning Model Thinks With Reinforcement Learning6 Mar
  6. 2025Towards Thinking-Optimal Scaling of Test-Time Compute for LLM Reasoning25 Feb
  7. 2025Revisiting the Test-Time Scaling of o1-like Models: Do they Truly Possess Test-Time Scaling Capabilities?17 Feb
  8. 2025TokenSkip: Controllable Chain-of-Thought Compression in LLMs17 Feb
  9. 2025CoT-Valve: Length-Compressible Chain-of-Thought Tuning13 Feb
  10. 2025s1: Simple test-time scaling31 Jan
  11. 2025Thoughts Are All Over the Place: On the Underthinking of o1-Like LLMs30 Jan
  12. 2024Efficiently Scaling LLM Reasoning with Certaindex30 Dec
  13. 2024Token-Budget-Aware LLM Reasoning24 Dec
  14. 2023Alphazero-like Tree-Search can Guide Large Language Model Decoding and Training29 Sep
  15. 2023Tree of Thoughts: Deliberate Problem Solving with Large Language Models17 May

Other references · 31

  1. 2024The Llama 3 Herd of Modelscited by 14,715
  2. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  3. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,330
  4. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,944
  5. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,708
  6. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,371
  7. 2024Qwen2.5 Technical Reportcited by 4,942
  8. 2022Large Language Models are Zero-Shot Reasonerscited by 4,374
  9. 2023Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Promptingcited by 1,122
  10. 2024Large Language Monkeys: Scaling Inference Compute with Repeated Samplingcited by 991
  11. 2023Measuring Faithfulness in Chain-of-Thought Reasoningcited by 703
  12. 2025Kimi K2: Open Agentic Intelligencecited by 567
19 more
  1. 2025Cognitive Behaviors that Enable Self-Improving Reasoners, or, Four Habits of Highly Effective STaRscited by 377
  2. 2025Chain of Draft: Thinking Faster by Writing Lesscited by 297
  3. 2025MathArena: Evaluating LLMs on Uncontaminated Math Competitionscited by 297
  4. 2025Thought Anchors: Which LLM Reasoning Steps Matter?cited by 129
  5. 2025Sketch-of-Thought: Efficient LLM Reasoning with Adaptive Cognitive-Inspired Sketchingcited by 117
  6. 2024Making Reasoning Matter: Measuring and Improving Faithfulness of Chain-of-Thought Reasoningcited by 110
  7. 2025Do Large Language Model Benchmarks Test Reliability?cited by 54
  8. 2025I Have Covered All the Bases Here: Interpreting Reasoning Features in Large Language Models via Sparse Autoencoderscited by 45
  9. 2025What Characterizes Effective Reasoning? Revisiting Length, Review, and Structure of CoTcited by 25
  10. 2025Can Aha Moments Be Fake? Towards Quantifying Decorative and True Thinking in Chain-of-Thoughtcited by 15
  11. 2026BibAgent: An Agentic Framework for Traceable Miscitation Detection in Scientific Literaturecited by 9
  12. 2026Position: Human-Centric AI Requires a Minimum Viable Level of Human Understandingcited by 6
  13. 2025A Language Anchor-Guided Method for Robust Noisy Domain Generalizationcited by 5
  14. 2026Traversal-as-Policy: Log-Distilled Gated Behavior Trees as Externalized, Verifiable Policies for Safe, Robust, and Efficient Agentscited by 5
  15. 2026WMF-AM: Probing LLM Working Memory via Depth-Parameterized Cumulative State Trackingcited by 4
  16. 2025TimePre: Bridging Accuracy, Efficiency, and Stability in Probabilistic Time-Series Forecastingcited by 3
  17. 2026Let the Abyss Stare Back Adaptive Falsification for Autonomous Scientific Discoverycited by 3
  18. 2025Unveiling the Latent Directions of Reflection in Large Language Modelscited by 2
  19. 2026AdaptFuse: Training-Free Sequential Preference Inference via Externalized Bayesian Computationcited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Understanding and Steering the Cognitive Behaviors of Reasoning Models at Test-Time
  2. 2025Thought calibration: Efficient and confident test-time scaling
  3. 2026When LLM Meets Tree Search: A Systematic View of Inference as Search in Large Language Models
  4. 2026TRACES: Tagging Reasoning Steps for Adaptive Cost-Efficient Early-Stopping
  5. 2025Reasoning on a Budget: A Survey of Adaptive and Controllable Test-Time Compute in LLMs
  6. 2026Thinking Traps in Long Chain-of-Thought: A Measurable Study and Trap-Aware Adaptive Restart
  7. 2026State of Thought Enables Endogenous Reasoning
  8. 2025ReflCtrl: Controlling LLM Reflection Efficiently via Representation Engineering