mlsanityData to 7 Oct 2026

Make Every Penny Count: Difficulty-Adaptive Self-Consistency for Cost-Efficient Reasoning

arXiv 2408.13457 ↗First posted 24 Aug 2024cs.CL cs.AIcited by 61

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

Abstract (arXiv)

Self-consistency (SC), a widely used decoding strategy for chain-of-thought reasoning, shows significant gains across various multi-step reasoning tasks but comes with a high cost due to multiple sampling with the preset size. Its variants, Adaptive self-consistency (ASC) and Early-stopping self-consistency (ESC), dynamically adjust the number of samples based on the posterior distribution of a set of pre-samples, reducing the cost of SC with minimal impact on performance. Both methods, however, do not exploit the prior information about question difficulty. It often results in unnecessary repeated sampling for easy questions that could be accurately answered with just one attempt, wasting resources. To tackle this problem, we propose Difficulty-Adaptive Self-Consistency (DSC), which leverages the difficulty information of batch queries from both prior and posterior perspectives to adaptively allocate inference resources, further reducing the overall cost of SC. To demonstrate the effectiveness of DSC, we conduct extensive experiments on three popular categories of reasoning tasks: arithmetic, commonsense and symbolic reasoning on six benchmarks. The empirical results show that DSC consistently surpasses the strong baseline ASC and ESC in terms of costs by a significant margin, while attaining comparable performances.

Key facts. Cited by 61 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 21 of its 34 references that are in the map. First posted on arXiv 24 Aug 2024.

Builds on

21 of 34 references.

Closest ancestors · 4

Same direction, newest first

  1. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters6 Aug
  2. 2024Integrate the Essence and Eliminate the Dross: Fine-Grained Self-Consistency for Free-Form Language Generation2 Jul
  3. 2024Escape Sky-high Cost: Early-stopping Self-Consistency for Multi-step Reasoning19 Jan
  4. 2023Let's Sample Step by Step: Adaptive-Consistency for Efficient Reasoning and Coding with LLMs19 May

Other references · 17

  1. 2021Training Verifiers to Solve Math Word Problemscited by 9,743
  2. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,601
  3. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,203
  4. 2022Large Language Models are Zero-Shot Reasonerscited by 4,323
  5. 2023Lost in the Middle: How Language Models Use Long Contextscited by 3,383
  6. 2023Sparks of Artificial General Intelligence: Early experiments with GPT-4cited by 2,579
  7. 2018CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledgecited by 1,793
  8. 2024Large Language Monkeys: Scaling Inference Compute with Repeated Samplingcited by 946
  9. 2021Are NLP Models really able to Solve Simple Math Word Problems?cited by 861
  10. 2021Did Aristotle Use a Laptop? A Question Answering Benchmark with Implicit Reasoning Strategiescited by 736
  11. 2016Solving General Arithmetic Word Problemscited by 353
  12. 2021A Diverse Corpus for Evaluating and Developing English Math Word Problem Solverscited by 314
5 more
  1. 2023Specializing Smaller Language Models towards Multi-Step Reasoningcited by 249
  2. 2023Batch Prompting: Efficient Inference with Large Language Model APIscited by 84
  3. 2024Large Language Models Can Self-Correct with Key Condition Verificationcited by 39
  4. 2024Multi-Task Inference: Can Large Language Models Follow Multiple Instructions at Once?cited by 28
  5. 2024CliqueParcel: An Approach For Batching LLM Prompts That Jointly Optimizes Efficiency And Faithfulnesscited by 12

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 5; 2025: 35; 2026: 21.

61 papers.

Sort:

By area:

  1. 2024Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMscited by 564
  2. 2025Stop Overthinking: A Survey on Efficient Reasoning for Large Language Modelscited by 456
  3. 2025A Survey on Test-Time Scaling in Large Language Models: What, How, Where, and How Well?cited by 215
  4. 2025Reasoning Models Know When They're Right: Probing Hidden States for Self-Verificationcited by 154
  5. 2025A Survey of Efficient Reasoning for Large Reasoning Models: Language, Multimodality, and Beyondcited by 131
  6. 2025Between Underthinking and Overthinking: An Empirical Study of Reasoning Length and correctness in LLMscited by 115
  7. 2025Confidence Improves Self-Consistency in LLMscited by 111
  8. 2025Efficient Reasoning Models: A Surveycited by 93
  9. 2025Efficient Inference for Large Reasoning Models: A Surveycited by 58
  10. 2025Harnessing the Reasoning Economy: A Survey of Efficient Reasoning for Large Language Modelscited by 35
  11. 2025Think or Not? Exploring Thinking Efficiency in Large Reasoning Models via an Information-Theoretic Lenscited by 32
  12. 2025Reasoning on a Budget: A Survey of Adaptive and Controllable Test-Time Compute in LLMscited by 31
38 more
  1. 2025Think Smarter not Harder: Adaptive Reasoning with Inference Aware Optimizationcited by 29
  2. 2025BudgetThinker: Empowering Budget-aware LLM Reasoning with Control Tokenscited by 27
  3. 2025The LLM Already Knows: Estimating LLM-Perceived Question Difficulty via Hidden Representationscited by 26
  4. 2024From Medprompt to o1: Exploration of Run-Time Strategies for Medical Challenge Problems and Beyondcited by 24
  5. 2025Route to Reason: Adaptive Routing for LLM and Reasoning Strategy Selectioncited by 23
  6. 2024Automatic Curriculum Expert Iteration for Reliable LLM Reasoningcited by 16
  7. 2025An Empirical Study of LLM Reasoning Ability Under Strict Output Length Constraintcited by 16
  8. 2025Generative AI Act II: Test Time Scaling Drives Cognition Engineeringcited by 14
  9. 2026Parallel-Probe: Towards Efficient Parallel Thinking via 2D Probingcited by 13
  10. 2025Adaptive GoGI-Skip: Coupling Goal-Gradient Importance with Dynamic Uncertainty for Efficient Reasoningcited by 12
  11. 2025Certified Self-Consistency: Statistical Guarantees and Test-Time Training for Reliable Reasoning in LLMscited by 12
  12. 2025Revisiting Self-Consistency from Dynamic Distributional Alignment Perspective on Answer Aggregationcited by 11
  13. 2024Reversal of Thought: Enhancing Large Language Models with Preference-Guided Reverse Reasoning Warm-upcited by 10
  14. 2025Exploring the System 1 Thinking Capability of Large Reasoning Modelscited by 10
  15. 2026VLA-ATTC: Adaptive Test-Time Compute for VLA Models with Relative Action Critic Modelcited by 9
  16. 2025AdaBoN: Adaptive Best-of-N Alignmentcited by 8
  17. 2025Strategic Scaling of Test-Time Compute: A Bandit Learning Approachcited by 8
  18. 2025C3PO: Optimized Large Language Model Cascades with Probabilistic Cost Constraints for Reasoningcited by 8
  19. 2025Optimal Self-Consistency for Efficient Reasoning with Large Language Modelscited by 8
  20. 2024Adapting While Learning: Grounding LLMs for Scientific Problems with Intelligent Tool Usage Adaptationcited by 7
  21. 2026Neural Chain-of-Thought Search: Searching the Optimal Reasoning Path to Enhance Large Language Modelscited by 7
  22. 2026Optimal Bayesian Stopping for Efficient Inference of Consistent LLM Answerscited by 7
  23. 2026LLMs Improving LLMs: Agentic Discovery for Test-Time Scalingcited by 7
  24. 2026DART: Draft-Agreement Routing for Training-Free Adaptive Thinking Budgets in Hybrid Reasoning Modelscited by 7
  25. 2025Speculative Decoding for Multi-Sample Inferencecited by 6
  26. 2025Interactive Learning for LLM Reasoningcited by 4
  27. 2025DEPO: Dual-Efficiency Preference Optimization for LLM Agentscited by 4
  28. 2025Adaptive Rectification Sampling for Test-Time Compute Scalingcited by 3
  29. 2025OptScale: Probabilistic Optimality for Inference-time Scalingcited by 3
  30. 2025LATTS: Locally Adaptive Test-Time Scalingcited by 3
  31. 2025Seer Self-Consistency: Advance Budget Estimation for Adaptive Test-Time Scalingcited by 3
  32. 2025Estimating problem difficulty without ground truth using Large Language Model comparisonscited by 3
  33. 2026Not All Turns Are Equally Hard: Adaptive Thinking Budgets For Efficient Multi-Turn Reasoning in Agentscited by 3
  34. 2026PETS: A Principled Framework Towards Optimal Trajectory Allocation for Efficient Test-Time Self-Consistencycited by 2
  35. 2026Resource Consumption Threats in Large Language Modelscited by 2
  36. 2026The Shadow Price of Reasoning: Economic Perspective on Optimal Budget Allocation for LLMscited by 2
  37. 2025PD3F: A Pluggable and Dynamic DoS-Defense Framework Against Resource Consumption Attacks Targeting Large Language Modelscited by 1
  38. 2025FEval-TTC: Fair Evaluation Protocol for Test-Time Computecited by 1

50 most cited of 61 All 61 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2026Is Discrete Difficulty Sufficient? Leveraging Continuous Difficulty for Efficient Self-Consistency in LLMs
  2. 2026Reliability-Aware Adaptive Self-Consistency for Efficient Sampling in LLM Reasoning
  3. 2024Reasoning Aware Self-Consistency: Leveraging Reasoning Paths for Efficient LLM Sampling
  4. 2026Breaking the Pre-Sampling Barrier: Activation-Informed Difficulty-Aware Self-Consistency
  5. 2026Self-Consistency via Marginal Sharpening
  6. 2025Self-Consistency Is Losing Its Edge: Diminishing Returns and Rising Costs in Modern LLMs
  7. 2023Universal Self-Consistency for Large Language Model Generation
  8. 2025Slim-SC: Thought Pruning for Efficient Scaling with Self-Consistency