mlsanityData to 14 Oct 2026

Reliable Chain-of-Thought via Prefix Consistency

arXiv 2605.07654 ↗First posted 8 May 2026stat.ML cs.CL cs.LGcited by 3

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

Abstract (arXiv)

Large Language Models often improve accuracy on reasoning tasks by sampling multiple Chain-of-Thought (CoT) traces and aggregating them with majority voting (MV), a test-time technique called self-consistency. When we truncate a CoT partway through and regenerate the remainder, we observe that traces with correct answers reproduce their original answer more often than traces with wrong answers. We use this difference as a reliability signal, prefix consistency, that weights each candidate answer by how often it reappears under regeneration. It requires no access to token log-probabilities or self-rating prompts. Across five reasoning models and four math and science benchmarks, prefix consistency is the best correctness predictor in most settings, and reweighting votes by it reaches Standard MV plateau accuracy at up to 21x fewer tokens (median 4.6x). Our code is available at https://github.com/naoto-iwase/prefix-consistency.

Key facts. Cited by 3 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 48 of its 51 references that are in the map. First posted on arXiv 8 May 2026.

Builds on

48 of 51 references.

Closest ancestors · 20

Same direction, newest first

  1. 2026The Path of Least Resistance: Guiding LLM Reasoning Trajectories with Prefix Consensus29 Jan
  2. 2025Nemotron 3 Nano: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning23 Dec
  3. 2025The Sequential Edge: Inverse-Entropy Voting Beats Parallel Self-Consistency at Matched Compute4 Nov
  4. 2025EAGer: Entropy-Aware GEneRation for Adaptive Inference-Time Scaling13 Oct
  5. 2025Deep Think with Confidence21 Aug
  6. 2025NVIDIA Nemotron Nano 2: An Accurate and Efficient Hybrid Mamba-Transformer Reasoning Model20 Aug
  7. 2025OptimalThinkingBench: Evaluating Over and Underthinking in LLMs18 Aug
  8. 2025gpt-oss-120b & gpt-oss-20b Model Card8 Aug
  9. 2025THOUGHTTERMINATOR: Benchmarking, Calibrating, and Mitigating Overthinking in Reasoning Models17 Apr
  10. 2025Reasoning Models Know When They're Right: Probing Hidden States for Self-Verification7 Apr
  11. 2025Scalable Best-of-N Selection for Large Language Models via Self-Certainty25 Feb
  12. 2025Inference-Time Computations for LLM Reasoning and Planning: A Benchmark and Insights18 Feb
8 more
  1. 2025Confidence Improves Self-Consistency in LLMs10 Feb
  2. 2025s1: Simple test-time scaling31 Jan
  3. 2025Rethinking External Slow-Thinking: From Snowball Errors to Probability of Correct Reasoning26 Jan
  4. 2024Path-Consistency with Prefix Enhancement for Efficient Inference in LLMs25 Aug
  5. 2024Escape Sky-high Cost: Early-stopping Self-Consistency for Multi-step Reasoning19 Jan
  6. 2023Let's Sample Step by Step: Adaptive-Consistency for Efficient Reasoning and Coding with LLMs19 May
  7. 2023Tree of Thoughts: Deliberate Problem Solving with Large Language Models17 May
  8. 2022Language Models (Mostly) Know What They Know11 Jul

Other references · 28

  1. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  2. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,708
  3. 2022Large Language Models are Zero-Shot Reasonerscited by 4,374
  4. 2023Let's Verify Step by Stepcited by 3,789
  5. 2022Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answeringcited by 1,919
  6. 2022Least-to-Most Prompting Enables Complex Reasoning in Large Language Modelscited by 1,407
  7. 2023Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMscited by 974
  8. 2022Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questionscited by 923
  9. 2023Graph of Thoughts: Solving Elaborate Problems with Large Language Modelscited by 886
  10. 2022Teaching Models to Express Their Uncertainty in Wordscited by 751
  11. 2023Measuring Faithfulness in Chain-of-Thought Reasoningcited by 703
  12. 2022Complexity-Based Prompting for Multi-Step Reasoningcited by 388
16 more
  1. 2025MathArena: Evaluating LLMs on Uncontaminated Math Competitionscited by 297
  2. 2025Reinforcement Learning with Verifiable Rewards Implicitly Incentivizes Correct Reasoning in Base LLMscited by 289
  3. 2024Fact-Checking the Output of Large Language Models via Token-Level Uncertainty Quantificationcited by 162
  4. 2026Ministral 3cited by 114
  5. 2023Knowledge-Driven CoT: Exploring Faithful Reasoning in LLMs for Knowledge-intensive Question Answeringcited by 97
  6. 2023T-SciQ: Teaching Multimodal Chain-of-Thought Reasoning via Mixed Large Language Model Signals for Science Question Answeringcited by 60
  7. 2026FrontierScience: Evaluating AI's Ability to Perform Expert-Level Scientific Taskscited by 47
  8. 2026Reasoning Theater: Disentangling Model Beliefs from Chain-of-Thoughtcited by 40
  9. 2025Benchmarking Reasoning Robustness in Large Language Modelscited by 30
  10. 2025Metastable Dynamics of Chain-of-Thought Reasoning: Provable Benefits of Search, RL and Distillationcited by 25
  11. 2025What Characterizes Effective Reasoning? Revisiting Length, Review, and Structure of CoTcited by 25
  12. 2025Sample, Don't Search: Rethinking Test-Time Alignment for Language Modelscited by 22
  13. 2024QUEST: Quality-Aware Metropolis-Hastings Sampling for Machine Translationcited by 16
  14. 2025Beyond the Last Answer: Your Reasoning Trace Uncovers More than You Thinkcited by 16
  15. 2025Robust Answers, Fragile Logic: Probing the Decoupling Hypothesis in LLM Reasoningcited by 11
  16. 2026Are Reasoning LLMs Robust to Interventions on Their Chain-of-Thought?cited by 6

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026Prefix-Denoising Consistency: Test-Time Verification for Diffusion Language Modelscited by 0
  2. 2026Sample Count Is Not Enough: Candidate-Generation Strategy Shapes the Energy and Performance of LLM Test-Time Scalingcited by 0
  3. 2026Self-Adapting Group of Experts for Multi-Agent Reasoningcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Optimal Self-Consistency for Efficient Reasoning with Large Language Models
  2. 2026Does Inference Scaling Improve Reasoning Faithfulness? A Multi-Model Analysis of Self-Consistency Tradeoffs
  3. 2025Ranked Voting based Self-Consistency of Large Language Models
  4. 2024Mirror-Consistency: Harnessing Inconsistency in Majority Voting
  5. 2026A Reference-Free Score for Detecting Silent Reasoning Failures in Large Language Models
  6. 2026Quantifying Consistency in LLM Logical Reasoning via Structural Uncertainty
  7. 2026Lie to Me: How Faithful Is Chain-of-Thought Reasoning in Reasoning Models?
  8. 2025Self-Consistency Is Losing Its Edge: Diminishing Returns and Rising Costs in Modern LLMs