mlsanityData to 14 Oct 2026

Mirror-Consistency: Harnessing Inconsistency in Majority Voting

arXiv 2410.10857 ↗First posted 7 Oct 2024cs.CL cs.AIcited by 17

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

Abstract (arXiv)

Self-Consistency, a widely-used decoding strategy, significantly boosts the reasoning capabilities of Large Language Models (LLMs). However, it depends on the plurality voting rule, which focuses on the most frequent answer while overlooking all other minority responses. These inconsistent minority views often illuminate areas of uncertainty within the model's generation process. To address this limitation, we present Mirror-Consistency, an enhancement of the standard Self-Consistency approach. Our method incorporates a 'reflective mirror' into the self-ensemble decoding process and enables LLMs to critically examine inconsistencies among multiple generations. Additionally, just as humans use the mirror to better understand themselves, we propose using Mirror-Consistency to enhance the sample-based confidence calibration methods, which helps to mitigate issues of overconfidence. Our experimental results demonstrate that Mirror-Consistency yields superior performance in both reasoning accuracy and confidence calibration compared to Self-Consistency.

Key facts. Cited by 17 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 35 of its 50 references that are in the map. First posted on arXiv 7 Oct 2024.

Builds on

35 of 50 references.

Closest ancestors · 10

Same direction, newest first

  1. 2024Self-Contrast: Better Reflection Through Inconsistent Solving Perspectives4 Jan
  2. 2023LLMs cannot find reasoning errors, but can correct them given the error location14 Nov
  3. 2023Large Language Models Cannot Self-Correct Reasoning Yet3 Oct
  4. 2023SCREWS: A Modular Framework for Reasoning with Revisions20 Sep
  5. 2023Automatically Correcting Large Language Models: Surveying the landscape of diverse self-correction strategies6 Aug
  6. 2023SelfCheck: Using LLMs to Zero-Shot Check Their Own Step-by-Step Reasoning1 Aug
  7. 2023REFINER: Reasoning Feedback on Intermediate Representations4 Apr
  8. 2023Self-Refine: Iterative Refinement with Self-Feedback30 Mar
  9. 2022Large Language Models Can Self-Improve20 Oct
  10. 2022Language Models (Mostly) Know What They Know11 Jul

Other references · 25

  1. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  2. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,708
  3. 2022PaLM: Scaling Language Modeling with Pathwayscited by 5,326
  4. 2017On Calibration of Modern Neural Networkscited by 4,960
  5. 2015Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learningcited by 4,448
  6. 2022Large Language Models are Zero-Shot Reasonerscited by 4,374
  7. 2023Toolformer: Language Models Can Teach Themselves to Use Toolscited by 4,182
  8. 2016Simple and Scalable Predictive Uncertainty Estimation using Deep Ensemblescited by 3,926
  9. 2018A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attackscited by 1,485
  10. 2023Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debatecited by 1,011
  11. 2023Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMscited by 974
  12. 2021Are NLP Models really able to Solve Simple Math Word Problems?cited by 878
13 more
  1. 2022Teaching Models to Express Their Uncertainty in Wordscited by 751
  2. 2021Did Aristotle Use a Laptop? A Question Answering Benchmark with Implicit Reasoning Strategiescited by 743
  3. 2023Visual ChatGPT: Talking, Drawing and Editing with Visual Foundation Modelscited by 668
  4. 2023Language Models can Solve Computer Taskscited by 372
  5. 2020Calibration of Pre-trained Transformerscited by 315
  6. 2023Faithful Chain-of-Thought Reasoningcited by 288
  7. 2020Reducing conversational agents' overconfidence through linguistic calibrationcited by 210
  8. 2023GPT-4 Doesn't Know It's Wrong: An Analysis of Iterative Prompting for Reasoning Problemscited by 89
  9. 2023Answering Questions by Meta-Reasoning over Multiple Chains of Thoughtcited by 87
  10. 2023Data-Copilot: Bridging Billions of Data and Humans with Autonomous Workflowcited by 76
  11. 2024Calibrating Large Language Models with Sample Consistencycited by 73
  12. 2023Self-Polish: Enhance Reasoning in Large Language Models via Problem Refinementcited by 43
  13. 2024Self-Consistency Boosts Calibration for Math Reasoningcited by 13

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 10; 2026: 7.

17 papers.

Sort:

By area:

  1. 2025The Majority is not always right: RL training for solution aggregationcited by 47
  2. 2025Stepwise Guided Policy Optimization: Coloring your Incorrect Reasoning in GRPOcited by 19
  3. 2025Beyond Consensus: Mitigating the Agreeableness Bias in LLM Judge Evaluationscited by 16
  4. 2025ReAgent: Reversible Multi-Agent Reasoning for Knowledge-Enhanced Multi-Hop QAcited by 13
  5. 2026What If Consensus Lies? Selective-Complementary Reinforcement Learning at Test Timecited by 12
  6. 2025Temporal Consistency for LLM Reasoning Process Error Identificationcited by 10
  7. 2025Self-Harmony: Learning to Harmonize Self-Supervision and Self-Play in Test-Time Reinforcement Learningcited by 10
  8. 2025Optimal Self-Consistency for Efficient Reasoning with Large Language Modelscited by 9
  9. 2025Think in Parallel, Answer as One: Logit Averaging for Open-Ended Reasoningcited by 6
  10. 2026Beyond Consensus: Trace-Level Synthesis in Mixture of Agentscited by 4
  11. 2025Multi-chain Graph Refinement and Selection for Reliable Reasoning in Large Language Modelscited by 1
  12. 2026CascadeMind at SemEval-2026 Task 4: A Hybrid Neuro-Symbolic Cascade for Narrative Similaritycited by 1
5 more
  1. 2025GradeSQL: Test-Time Inference with Outcome Reward Models for Text-to-SQL Generation from Large Language Modelscited by 0
  2. 2026Discovering Novel LLM Experts via Task-Capability Coevolutioncited by 0
  3. 2026Neuro-Symbolic AI for LEED compliance: Document-Centric Benchmarking, Deterministic Numeric Checking, and When Multimodal Hurtscited by 0
  4. 2026Test-Time Augmentation for LLMs: When Input Diversity Beats Output Diversity at Matched Computecited by 0
  5. 2026Is This Your Final Answer? Cross-Contextual Consistency as a Measure of LLM Credibilitycited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Confidence Improves Self-Consistency in LLMs
  2. 2025Latent Self-Consistency for Reliable Majority-Set Selection in Short- and Long-Answer Reasoning
  3. 2024Integrative Decoding: Improve Factuality via Implicit Self-consistency
  4. 2025Certified Self-Consistency: Statistical Guarantees and Test-Time Training for Reliable Reasoning in LLMs
  5. 2024Internal Consistency and Self-Feedback in Large Language Models: A Survey
  6. 2025Self-Consistency Is Losing Its Edge: Diminishing Returns and Rising Costs in Modern LLMs
  7. 2026Does Inference Scaling Improve Reasoning Faithfulness? A Multi-Model Analysis of Self-Consistency Tradeoffs
  8. 2026Boosting Self-Consistency with Ranking