mlsanityData to 14 Oct 2026

PaceReasoner: Label-Free Learning to Answer Briefly from Long Reasoning

arXiv 2610.17900 ↗First posted 13 Oct 2026cs.AI cs.LGcited by 0

On the map: LLM reasoning → Test-time compute and efficient thinking also fits: Reinforcement learning for reasoning

Abstract (arXiv)

Extended chain-of-thought reasoning improves problem-solving performance but increases inference cost and latency, motivating more accurate answers with shorter reasoning. We introduce PaceReasoner, a label-free test-time reinforcement learning framework that uses a model's LONG responses to improve its SHORT-mode accuracy. Both modes share a trainable policy, creating a need for reliable supervision, sufficient SHORT training samples, and continued LONG responses to guide learning. PaceReasoner combines all-response and LONG-only answer consensus through progressive, confidence-weighted supervision. A behavior reward gradually shifts from encouraging SHORT exploration to rewarding agreement with LONG answers, while a ratio controller regulates the balance between modes. Across three model backbones and four benchmarks, PaceReasoner improves average SHORT accuracy over the baselines by 7.4--13.5 percentage points, with fewer than 25 reasoning tokens per response on average. Ablations support the key components, and cross-dataset results demonstrate transfer beyond the adaptation set.

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 18 of its 28 references that are in the map. First posted on arXiv 13 Oct 2026.

Builds on

18 of 28 references.

Closest ancestors · 5

Same direction, newest first

  1. 2025Steering LLM Thinking with Budget Guidance16 Jun
  2. 2025AdaptThink: Reasoning Models Can Learn When to Think19 May
  3. 2025Thinkless: LLM Learns When to Think19 May
  4. 2025L1: Controlling How Long A Reasoning Model Thinks With Reinforcement Learning6 Mar
  5. 2025CoT-Valve: Length-Compressible Chain-of-Thought Tuning13 Feb

Other references · 13

  1. 2024The Llama 3 Herd of Modelscited by 14,715
  2. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  3. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 9,379
  4. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,371
  5. 2023GPQA: A Graduate-Level Google-Proof Q&A Benchmarkcited by 2,879
  6. 2018CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledgecited by 1,822
  7. 2023On-Policy Distillation of Language Models: Learning from Self-Generated Mistakescited by 939
  8. 2025TTRL: Test-Time Reinforcement Learningcited by 275
  9. 2024From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Stepcited by 235
  10. 2025Learning to Reason without External Rewardscited by 210
  11. 2026CRISP: Compressed Reasoning via Iterative Self-Policy Distillationcited by 58
  12. 2026On-Policy Self-Distillation without Any Supervisioncited by 17
1 more
  1. 2026Masked Distillation: Internalizing the Chain-of-Thought in Language Modelscited by 3

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Consensus as Privileged Context for Label-Free Self-Distillation
  2. 2026Correct, Concise and Complete: Multi-stage Training For Adaptive Reasoning
  3. 2026CoVerRL: Breaking the Consensus Trap in Label-Free Reasoning via Generator-Verifier Co-Evolution
  4. 2026Stepwise Penalization for Length-Efficient Chain-of-Thought Reasoning
  5. 2025Rewarding the Journey, Not Just the Destination: A Composite Path and Answer Self-Scoring Reward Mechanism for Test-Time Reinforcement Learning
  6. 2026Agentic Chain-of-Thought Steering for Efficient and Controllable LLM Reasoning
  7. 2026LEAD: Length-Efficient Adaptive and Dynamic Reasoning for Large Language Models
  8. 2025Dissecting Long-Chain-of-Thought Reasoning Models: An Empirical Study