mlsanityData to 7 Oct 2026

ESTAR: Early-Stopping Token-Aware Reasoning For Efficient Inference

arXiv 2602.10004 ↗First posted 10 Feb 2026cs.AIcited by 3

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

Abstract (arXiv)

Large reasoning models (LRMs) achieve state-of-the-art performance by generating long chains-of-thought, but often waste computation on redundant reasoning after the correct answer has already been reached. We introduce Early-Stopping for Token-Aware Reasoning (ESTAR), which detects and reduces such reasoning redundancy to improve efficiency without sacrificing accuracy. Our method combines (i) a trajectory-based classifier that identifies when reasoning can be safely stopped, (ii) supervised fine-tuning to teach LRMs to propose self-generated <stop> signals, and (iii) <stop>-aware reinforcement learning that truncates rollouts at self-generated stop points with compute-aware rewards. Experiments on four reasoning datasets show that ESTAR reduces reasoning length by about 3.7x (from 4,799 to 1,290) while preserving accuracy (74.9% vs. 74.2%), with strong cross-domain generalization. These results highlight early stopping as a simple yet powerful mechanism for improving reasoning efficiency in LRMs.

Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 35 of its 42 references that are in the map. First posted on arXiv 10 Feb 2026.

Builds on

35 of 42 references.

Closest ancestors · 20

Same direction, newest first

  1. 2025The Evolution of Thought: Tracking LLM Overthinking via Reasoning Dynamics Analysis25 Aug
  2. 2025ConciseHint: Boosting Efficient Reasoning via Continuous Concise Hints during Generation23 Jun
  3. 2025On Reasoning Strength Planning in Large Reasoning Models10 Jun
  4. 2025Just Enough Thinking: Efficient Reasoning with Adaptive Length Penalties Reinforcement Learning5 Jun
  5. 2025CP-Router: An Uncertainty-Aware Router Between LLM and LRM26 May
  6. 2025Think Only When You Need with Large Hybrid-Reasoning Models20 May
  7. 2025FlashThink: An Early Exit Method For Efficient Reasoning20 May
  8. 2025AdaptThink: Reasoning Models Can Learn When to Think19 May
  9. 2025Thinkless: LLM Learns When to Think19 May
  10. 2025AdaCoT: Pareto-Optimal Adaptive Chain-of-Thought Triggering via Reinforcement Learning17 May
  11. 2025Dynamic Early Exit in Reasoning Models22 Apr
  12. 2025Reasoning Models Can Be Effective Without Thinking14 Apr
8 more
  1. 2025Reasoning Models Know When They're Right: Probing Hidden States for Self-Verification7 Apr
  2. 2025Concise Reasoning via Reinforcement Learning7 Apr
  3. 2025ThinkPrune: Pruning Long Chain-of-Thought of LLMs via Reinforcement Learning2 Apr
  4. 2025A Survey of Efficient Reasoning for Large Reasoning Models: Language, Multimodality, and Beyond27 Mar
  5. 2025Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models20 Mar
  6. 2025DAST: Difficulty-Adaptive Slow-Thinking for Large Reasoning Models6 Mar
  7. 2025Demystifying Long Chain-of-Thought Reasoning in LLMs5 Feb
  8. 2025s1: Simple test-time scaling31 Jan

Other references · 15

  1. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  2. 2025Qwen3 Technical Reportcited by 8,195
  3. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,203
  4. 2023GPQA: A Graduate-Level Google-Proof Q&A Benchmarkcited by 2,748
  5. 2020What Disease does this Patient Have? A Large-scale Open Domain Question Answering Dataset from Medical Examscited by 1,324
  6. 2025From System 1 to System 2: A Survey of Reasoning Large Language Modelscited by 319
  7. 2025Chain of Draft: Thinking Faster by Writing Lesscited by 290
  8. 2025Training Language Models to Reason Efficientlycited by 242
  9. 2025OverThink: Slowdown Attacks on Reasoning LLMscited by 94
  10. 2025S-GRPO: Early Exit via Reinforcement Learning in Reasoning Modelscited by 87
  11. 2025Sample More to Think Less: Group Filtered Policy Optimization for Concise Reasoningcited by 75
  12. 2025MedicalAgentsBench for Complex Medical Reasoning: Comparing Internalized Reasoning Models versus Externalized Agent-based Frameworkscited by 38
3 more
  1. 2025Think When You Need: Self-Adaptive Chain-of-Thought Learningcited by 37
  2. 2025Harnessing Chain-of-Thought Metadata for Task Routing and Adversarial Prompt Detectioncited by 5
  3. 2025Rethinking Reasoning Quality in Large Language Models through Enhanced Chain-of-Thought via RLcited by 5

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026ESPO: Early-Stopping Proximal Policy Optimizationcited by 1
  2. 2026The Language-Energy Divide: Measuring Energy Costs of Multilingual LLM Inferencecited by 0
  3. 2026Settle: Learning When to Stop Reasoningcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Beyond Token Length: Step Pruner for Efficient and Accurate Reasoning in Large Language Models
  2. 2025Stop When Enough: Adaptive Early-Stopping for Chain-of-Thought Reasoning
  3. 2026Stop When Reasoning Converges: Semantic-Preserving Early Exit for Reasoning Models
  4. 2026Efficient Reasoning via Constrained Optimization in Latent Space
  5. 2025Not All Tokens Matter: Towards Efficient LLM Reasoning via Token Significance in Reinforcement Learning
  6. 2025Learning When to Stop: Adaptive Latent Reasoning via Reinforcement Learning
  7. 2026Learning to Stop without Learning to Stop: Self-Supervised Confidence Training Improves Reasoning Efficiency
  8. 2026Stop When Further Reasoning Won't Help: Attention-State Adaptive Generation in Reasoning Models