mlsanityData to 14 Oct 2026

Amplified Does Not Mean Predictive: Reasoning Behaviors in Thinking Models

arXiv 2608.13760 ↗First posted 13 Aug 2026cs.CL cs.AI cs.CVcited by 1

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

Abstract (arXiv)

Which reasoning behaviors are associated with correct answers in reasoning models, and does reasoning-oriented training amplify those behaviors? This distinction is important because reasoning-oriented training can make traces look more deliberative without amplifying the behaviors most tied to model correctness. We quantify this mismatch with Behavioral Lift, a metric that measures how much correctness changes when a behavior is present versus absent in a model's reasoning trace. Across 15 models and 6 benchmarks spanning text-only and vision-language reasoning, we annotate 15,282 traces with a taxonomy whose core behaviors are defined for both LLM and VLM traces. We find evidence for an Amplification-Lift Gap, in which thinking models strongly amplify self-correction, hypothesis testing, and uncertainty acknowledgment, while the highest-lift behaviors are confidence calibration, knowledge alignment, and self-awareness. Confidence calibration is among the strongest positive signals of correctness in both modalities, yet is barely amplified; uncertainty acknowledgment is amplified by 3--7×, yet is weakly or negatively associated with correctness. We find that reasoning-oriented training does not preferentially amplify the highest-Lift behaviors, motivating process-level objectives that reward calibrated and grounded reasoning rather than surface form alone.

Key facts. Cited by 1 paper 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 27 of its 32 references that are in the map. First posted on arXiv 13 Aug 2026.

Builds on

27 of 32 references.

Closest ancestors · 3

Same direction, newest first

  1. 2025Reasoning Models Can Be Effective Without Thinking14 Apr
  2. 2025Demystifying Long Chain-of-Thought Reasoning in LLMs5 Feb
  3. 2025Thoughts Are All Over the Place: On the Underthinking of o1-Like LLMs30 Jan

Other references · 24

  1. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  2. 2025Qwen3 Technical Reportcited by 8,820
  3. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,371
  4. 2023Let's Verify Step by Stepcited by 3,789
  5. 2023MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGIcited by 2,080
  6. 2024MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmarkcited by 1,848
  7. 2023MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contextscited by 1,614
  8. 2025Kimi k1.5: Scaling Reinforcement Learning with LLMscited by 1,210
  9. 2025Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?cited by 982
  10. 2025SFT Memorizes, RL Generalizes: A Comparative Study of Foundation Model Post-trainingcited by 672
  11. 2025Reasoning Models Don't Always Say What They Thinkcited by 415
  12. 2025Monitoring Reasoning Models for Misbehavior and the Risks of Promoting Obfuscationcited by 384
12 more
  1. 2025Cognitive Behaviors that Enable Self-Improving Reasoners, or, Four Habits of Highly Effective STaRscited by 377
  2. 2024ProcessBench: Identifying Process Errors in Mathematical Reasoningcited by 292
  3. 2025Training Language Models to Reason Efficientlycited by 247
  4. 2026Why Does Self-Distillation (Sometimes) Degrade the Reasoning Capability of LLMs?cited by 103
  5. 2020LogiQA: A Challenge Dataset for Machine Reading Comprehension with Logical Reasoningcited by 69
  6. 2025On the Interplay of Pre-Training, Mid-Training, and RL on Reasoning Language Modelscited by 69
  7. 2025ImpossibleBench: Measuring LLMs' Propensity of Exploiting Test Casescited by 65
  8. 2025VisualPuzzles: Decoupling Multimodal Reasoning Evaluation from Domain Knowledgecited by 55
  9. 2026Reasoning Theater: Disentangling Model Beliefs from Chain-of-Thoughtcited by 40
  10. 2025Cognitive Foundations for Reasoning and Their Manifestation in LLMscited by 26
  11. 2025The CoT Encyclopedia: Analyzing, Predicting, and Controlling how a Reasoning Model will Thinkcited by 9
  12. 2026PRISM: A Dual View of LLM Reasoning through Semantic Flow and Latent Computationcited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Hallucination Across the Reasoning Lifecycle: Interface Visibility, Causal Evidence, and Release Control in Large Reasoning Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Thinking Out Loud: Do Reasoning Models Know When They're Right?
  2. 2025Reasoning about Uncertainty: Do Reasoning Models Know When They Don't Know?
  3. 2025Reasoning Models Better Express Their Confidence
  4. 2026The Illusion of Insight in Reasoning Models
  5. 2026Nudgeability: Reasoning Models Follow Confidence Signals Without Tracking Their Own Competence
  6. 2026Calibration Drift Under Reasoning: How Chain-of-Thought Budgets Induce Overconfidence in Large Language Models
  7. 2026The Cost of Reasoning: Chain-of-Thought Induces Overconfidence in Vision-Language Models
  8. 2025Large Reasoning Models are not thinking straight: on the unreliability of thinking trajectories