mlsanityData to 14 Oct 2026

When More Thinking Hurts: Overthinking in LLM Test-Time Compute Scaling

arXiv 2604.10739 ↗First posted 12 Apr 2026cs.AIcited by 16

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

Abstract (arXiv)

Scaling test-time compute through extended chains of thought has become a dominant paradigm for improving large language model reasoning. However, existing research implicitly assumes that longer thinking always yields better results. This assumption remains largely unexamined. We systematically investigate how the marginal utility of additional reasoning tokens changes as compute budgets increase. We find that marginal returns diminish substantially at higher budgets and that models exhibit ``overthinking'', where extended reasoning is associated with abandoning previously correct answers. Furthermore, we show that optimal thinking length varies across problem difficulty, suggesting that uniform compute allocation is suboptimal. Our cost-aware evaluation framework reveals that stopping at moderate budgets can reduce computation significantly while maintaining comparable accuracy.

Key facts. Cited by 16 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 14 of its 26 references that are in the map. First posted on arXiv 12 Apr 2026.

Builds on

14 of 26 references.

Closest ancestors · 7

Same direction, newest first

  1. 2025Do LLMs Really Need 10+ Thoughts for "Find the Time 1000 Days Later"? Towards Structural Understanding of LLM Overthinking9 Oct
  2. 2025Does Thinking More always Help? Mirage of Test-Time Scaling in Reasoning Models4 Jun
  3. 2025Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models20 Mar
  4. 2025Is That Your Final Answer? Test-Time Scaling Improves Selective Question Answering19 Feb
  5. 2025s1: Simple test-time scaling31 Jan
  6. 2024Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs30 Dec
  7. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters6 Aug

Other references · 7

  1. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,330
  2. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,371
  3. 2022Fast Inference from Transformers via Speculative Decodingcited by 1,653
  4. 2025Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Modelscited by 384
  5. 2025When More is Less: Understanding Chain-of-Thought Length in LLMscited by 183
  6. 2020The Right Tool for the Job: Matching Model and Instance Complexitiescited by 126
  7. 2025Do LLMs Overthink Basic Math Reasoning? Benchmarking the Accuracy-Efficiency Tradeoff in Language Modelscited by 5

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 16.

16 papers.

Sort:

By area:

  1. 2026SCOPE-RL: Optimizing Reasoning Paths Before and After Successcited by 1
  2. 2026LLM-based Agents for Forecasting and Prediction: Methods, Training, Evaluation, and Applicationscited by 1
  3. 2026It's the Problem, Not the Path: Budget and Difficulty Confounds in LLM Reasoning Trajectoriescited by 1
  4. 2026Know When to Trust the Skill: Delayed Appraisal and Epistemic Vigilance for Single-Agent LLMscited by 0
  5. 2026State Stream Transformer (SST) V2: Parallel Training of Nonlinear Recurrence for Latent Space Reasoningcited by 0
  6. 2026CRANE: Constrained Reasoning Injection for Code Agents via Nullspace Editingcited by 0
  7. 2026Off-the-Shelf LLMs as Process Scorers: Training-Free Alternative to PRMs for Mathematical Reasoningcited by 0
  8. 2026Projecting the Emerging Mindset of SWE Agent by Launching a Wild Code Understanding Journeycited by 0
  9. 2026Solver-Hard Is Not Model-Hard: A Hardness-Controlled Diagnostic for LLM Constraint Reasoningcited by 0
  10. 2026Don't Overthink, Don't Underthink: Toward Adaptive Reasoning in Agentic AIcited by 0
  11. 2026InSituMeasure: Probing Situated Measurement Grounding in Industrial Scenes with Multimodal Large Language Modelscited by 0
  12. 2026LLM Parkinsonism: Executive-Control Failure, Token-Inefficient Persistence, and an Uncertainty-Aware Global Executive Control Architecture for Autonomous Language-Model Agentscited by 0
4 more
  1. 2026MetaCtrl: Your Large Language Models Can Reason Better and More Concisely with a Metacognitive Controllercited by 0
  2. 2026Knowing the Rules, Applying the Rules: Evaluating Language Models on Traditional Chinese Bazicited by 0
  3. 2026Whose turn is it? Context-guided model routing for stateful agentscited by 0
  4. 2026RCCM: Diagnosing Gains and Harms in Reasoning-Prompt Transfercited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Towards Thinking-Optimal Scaling of Test-Time Compute for LLM Reasoning
  2. 2025ParaThinker: Native Parallel Thinking as a New Paradigm to Scale LLM Test-time Compute
  3. 2025The Evolution of Thought: Tracking LLM Overthinking via Reasoning Dynamics Analysis
  4. 2025Reasoning on a Budget: A Survey of Adaptive and Controllable Test-Time Compute in LLMs
  5. 2025The Art of Scaling Test-Time Compute for Large Language Models
  6. 2025The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity
  7. 2025Thinking Short and Right Over Thinking Long: Serving LLM Reasoning Efficiently and Accurately
  8. 2025Don't Overthink it. Preferring Shorter Thinking Chains for Improved LLM Reasoning