mlsanityData to 7 Oct 2026

When to Trust the Cheap Check: Weak and Strong Verification for Reasoning

arXiv 2602.17633 ↗First posted 19 Feb 2026cs.LG cs.AI stat.MLcited by 5

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

Abstract (arXiv)

Reasoning with LLMs increasingly unfolds inside a broader verification loop. Internally, systems use cheap checks, such as self-consistency or proxy rewards, which we call weak verification. Externally, users inspect outputs and steer the model through feedback until results are trustworthy, which we call strong verification. These signals differ sharply in cost and reliability: strong verification can establish trust but is resource-intensive, while weak verification is fast and scalable but noisy and imperfect. We formalize this tension through weak--strong verification policies, which decide when to accept or reject based on weak verification and when to defer to strong verification. We introduce metrics capturing incorrect acceptance, incorrect rejection, and strong-verification frequency. Over population, we show that optimal policies admit a two-threshold structure and that calibration and sharpness govern the value of weak verifiers. Building on this, we develop an online algorithm that provably controls acceptance and rejection errors without assumptions on the query stream, the language model, or the weak verifier.

Key facts. Cited by 5 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 29 of its 35 references that are in the map. First posted on arXiv 19 Feb 2026.

Builds on

29 of 35 references.

Closest ancestors · 2

Same direction, newest first

  1. 2023Tree of Thoughts: Deliberate Problem Solving with Large Language Models17 May
  2. 2023Self-Evaluation Guided Beam Search for Reasoning1 May

Other references · 27

  1. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,302
  2. 2021Training Verifiers to Solve Math Word Problemscited by 9,743
  3. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  4. 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,355
  5. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,203
  6. 2023Let's Verify Step by Stepcited by 3,598
  7. 2023G-Eval: NLG Evaluation using GPT-4 with Better Human Alignmentcited by 2,205
  8. 2023Math-Shepherd: Verify and Reinforce LLMs Step-by-step without Human Annotationscited by 895
  9. 2017Selective Classification for Deep Neural Networkscited by 799
  10. 2024Generative Verifiers: Reward Modeling as Next-Token Predictioncited by 465
  11. 2021Adaptive Conformal Inference Under Distribution Shiftcited by 360
  12. 2025Reinforcement Learning for Reasoning in Large Language Models with One Training Examplecited by 288
15 more
  1. 2020Consistent Estimators for Learning to Defer to an Expertcited by 262
  2. 2023ARES: An Automated Evaluation Framework for Retrieval-Augmented Generation Systemscited by 259
  3. 2024MiniCheck: Efficient Fact-Checking of LLMs on Grounding Documentscited by 251
  4. 2024Monte Carlo Tree Search Boosts Reasoning via Iterative Preference Learningcited by 238
  5. 2022Speculative Decoding: Exploiting Speculative Execution for Accelerating Seq2seq Generationcited by 169
  6. 2023HelpSteer: Multi-attribute Helpfulness Dataset for SteerLMcited by 140
  7. 2024Fast Best-of-N Decoding via Speculative Rejectioncited by 112
  8. 2023Conformal PID Control for Time Series Predictioncited by 85
  9. 2021Differentiable Learning Under Triagecited by 74
  10. 2022Calibrated Learning to Defer with One-vs-All Classifierscited by 68
  11. 2023Who Should Predict? Exact Algorithms For Learning to Defer to Humanscited by 68
  12. 2024Archon: An Architecture Search Framework for Inference-Time Techniquescited by 59
  13. 2025The Relationship between No-Regret Learning and Online Conformal Predictioncited by 13
  14. 2021Online Selective Classification with Limited Feedbackcited by 8
  15. 2025Verdict: A Library for Scaling Judge-Time Computecited by 7

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 5.

5 papers.

  1. 2026Calibrate-Then-Delegate: Safety Monitoring with Risk and Budget Guarantees via Model Cascadescited by 3
  2. 2026Adaptive Generate-Rank-Verify: Inference-Time Search with Costly Verificationcited by 1
  3. 2026Conformal Risk-Averse Decision Making with Action Conditional Guaranteecited by 1
  4. 2026Strategic Decision Support for AI Agentscited by 1
  5. 2026More Rejective, Not More Discriminative: The Unit of Verification in Pre-Execution LLM Oversightcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Self-Verification Dilemma: Experience-Driven Suppression of Overused Checking in LLM Reasoning
  2. 2026Grading the Graders: Verification Autonomy Levels (L0-L5) for LLM Reasoning
  3. 2026Measuring Weak-to-Strong Legibility of Reasoning Models
  4. 2026Trust but Verify: Prover-Verifier Deliberation for Selective LLM Prediction
  5. 2025When Does Verification Pay Off? A Closer Look at LLMs as Solution Verifiers
  6. 2026Agentic Systems as Boosting Weak Reasoning Models
  7. 2026Viva La Vida: Verification and Accumulation Failures in Multi-Agent Proof Search
  8. 2026Project Aletheia: Verifier-Guided Distillation of Backtracking for Small Language Models