mlsanityData to 14 Oct 2026

Contrastive Decoding Improves Reasoning in Large Language Models

arXiv 2309.09117 ↗First posted 17 Sep 2023cs.CL cs.AIcited by 74

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

Abstract (arXiv)

We demonstrate that Contrastive Decoding -- a simple, computationally light, and training-free text generation method proposed by Li et al 2022 -- achieves large out-of-the-box improvements over greedy decoding on a variety of reasoning tasks. Originally shown to improve the perceived quality of long-form text generation, Contrastive Decoding searches for strings that maximize a weighted difference in likelihood between strong and weak models. We show that Contrastive Decoding leads LLaMA-65B to outperform LLaMA 2, GPT-3.5 and PaLM 2-L on the HellaSwag commonsense reasoning benchmark, and to outperform LLaMA 2, GPT-3.5 and PaLM-540B on the GSM8K math word reasoning benchmark, in addition to improvements on a collection of other tasks. Analysis suggests that Contrastive Decoding improves over existing methods by preventing some abstract reasoning errors, as well as by avoiding simpler modes such as copying sections of the input during chain-of-thought. Overall, Contrastive Decoding outperforms nucleus sampling for long-form generation and greedy decoding for reasoning tasks, making it a powerful general purpose method for generating text from language models.

Key facts. Cited by 74 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 30 of its 35 references that are in the map. First posted on arXiv 17 Sep 2023.

Builds on

30 of 35 references.

References · 30

  1. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 14,132
  2. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  3. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  4. 2020Measuring Massive Multitask Language Understandingcited by 7,963
  5. 2020Scaling Laws for Neural Language Modelscited by 6,586
  6. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,708
  7. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,371
  8. 2018Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challengecited by 4,886
  9. 2019HellaSwag: Can a Machine Really Finish Your Sentence?cited by 4,100
  10. 2017TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehensioncited by 2,955
  11. 2019The Curious Case of Neural Text Degenerationcited by 2,916
  12. 2019PIQA: Reasoning about Physical Commonsense in Natural Languagecited by 2,883
18 more
  1. 2019BoolQ: Exploring the Surprising Difficulty of Natural Yes/No Questionscited by 2,183
  2. 2018Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answeringcited by 2,164
  3. 2018CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledgecited by 1,822
  4. 2018Hierarchical Neural Story Generationcited by 1,193
  5. 2021Are NLP Models really able to Solve Simple Math Word Problems?cited by 878
  6. 2021Did Aristotle Use a Laptop? A Question Answering Benchmark with Implicit Reasoning Strategiescited by 743
  7. 2017Program Induction by Rationale Generation : Learning to Solve and Explain Algebraic Word Problemscited by 650
  8. 2022Contrastive Decoding: Open-ended Text Generation as Optimizationcited by 609
  9. 2021DExperts: Decoding-Time Controlled Text Generation with Experts and Anti-Expertscited by 441
  10. 2021FUDGE: Controlled Text Generation With Future Discriminatorscited by 376
  11. 2023DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Modelscited by 376
  12. 2021A Diverse Corpus for Evaluating and Developing English Math Word Problem Solverscited by 318
  13. 2022Reasoning with Language Model Prompting: A Surveycited by 252
  14. 2022Towards Understanding Chain-of-Thought Prompting: An Empirical Study of What Matterscited by 222
  15. 2022ROSCOE: A Suite of Metrics for Scoring Step-by-Step Reasoningcited by 177
  16. 2022Locally Typical Samplingcited by 111
  17. 2023Chain-of-Thought Hub: A Continuous Effort to Measure Large Language Models' Reasoning Performancecited by 102
  18. 2023Surfacing Biases in Large Language Models using Contrastive Input Decodingcited by 11

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 4; 2024: 24; 2025: 17; 2026: 29.

74 papers.

Sort:

By area:

  1. 2023Mitigating Object Hallucinations in Large Vision-Language Models through Visual Contrastive Decodingcited by 688
  2. 2024Chain-of-Thought Reasoning Without Promptingcited by 210
  3. 2024Critical Tokens Matter: Token-Level Contrastive Estimation Enhances LLM's Reasoning Capabilitycited by 83
  4. 2024Mitigating Object Hallucinations in Large Vision-Language Models with Assembly of Global and Local Attentioncited by 74
  5. 2024Interpretable Contrastive Monte Carlo Tree Search Reasoningcited by 64
  6. 2023The Consensus Game: Language Model Generation via Equilibrium Searchcited by 36
  7. 2024mCoT: Multilingual Instruction Tuning for Reasoning Consistency in Language Modelscited by 33
  8. 2025Spotlight Your Instructions: Instruction-following with Dynamic Attention Steeringcited by 21
  9. 2023Supervised Knowledge Makes Large Language Models Better In-context Learnerscited by 20
  10. 2024Seeing the Image: Prioritizing Visual Correlation by Contrastive Alignmentcited by 20
  11. 2024Unchosen Experts Can Contribute Too: Unleashing MoE Models' Power by Self-Contrastcited by 19
  12. 2024Entropy-Based Decoding for Retrieval-Augmented Large Language Modelscited by 19
38 more
  1. 2024Distillation Contrastive Decoding: Improving LLMs Reasoning with Contrastive Decoding and Distillationcited by 16
  2. 2024Search, Verify and Feedback: Towards Next Generation Post-training Paradigm of Foundation Models via Verifier Engineeringcited by 16
  3. 2024Learning From Correctness Without Prompting Makes LLM Efficient Reasonercited by 15
  4. 2024Rapid Optimization for Jailbreaking LLMs via Subconscious Exploitation and Echopraxiacited by 14
  5. 2024Fast and Slow Generating: An Empirical Study on Large and Small Language Models Collaborative Decodingcited by 13
  6. 2024Pseudo-RIS: Distinctive Pseudo-supervision Generation for Referring Image Segmentationcited by 13
  7. 2024VaLiD: Mitigating the Hallucination of Large Vision Language Models by Visual Layer Fusion Contrastive Decodingcited by 12
  8. 2024LLM2: Let Large Language Models Harness System 2 Reasoningcited by 12
  9. 2025Fast Large Language Model Collaborative Decoding via Speculationcited by 12
  10. 2023Context-aware Decoding Reduces Hallucination in Query-focused Summarizationcited by 11
  11. 2025TruthPrInt: Mitigating Large Vision-Language Models Object Hallucination Via Latent Truthful-Guided Pre-Interventioncited by 11
  12. 2025On the Consistency of Multilingual Context Utilization in Retrieval-Augmented Generationcited by 10
  13. 2025UCD: Unlearning in LLMs via Contrastive Decodingcited by 9
  14. 2024Towards Building Specialized Generalist AI with System 1 and System 2 Fusioncited by 8
  15. 2024Improving Large Language Model (LLM) fidelity through context-aware grounding: A systematic approach to reliability and veracitycited by 8
  16. 2024Integrative Decoding: Improve Factuality via Implicit Self-consistencycited by 8
  17. 2025UniDetox: Universal Detoxification of Large Language Models via Dataset Distillationcited by 8
  18. 2025ProxyThinker: Test-Time Guidance through Small Visual Reasonerscited by 8
  19. 2026ThinkOmni: Lifting Textual Reasoning to Omni-modal Scenarios via Guidance Decodingcited by 7
  20. 2025Jakiro: Boosting Speculative Decoding with Decoupled Multi-Head via MoEcited by 6
  21. 2026Thinking by Subtraction: Confidence-Driven Contrastive Decoding for LLM Reasoningcited by 6
  22. 2026Neglected Free Lunch from Post-training: Progress Advantage for LLM Agentscited by 6
  23. 2025RAST: Reasoning Activation in LLMs via Small-model Transfercited by 5
  24. 2025DecoRTL: A Run-time Decoding Framework for RTL Code Generation with LLMscited by 5
  25. 2026Self-Improvement of Large Language Models: A Technical Overview and Future Outlookcited by 5
  26. 2025Advancing Decoding Strategies: Enhancements in Locally Typical Sampling for LLMscited by 4
  27. 2025CROP: Circuit Retrieval and Optimization with Parameter Guidance using LLMscited by 3
  28. 2025LightReasoner: Can Small Language Models Teach Large Language Models Reasoning?cited by 3
  29. 2026NoLan: Mitigating Object Hallucinations in Large Vision-Language Models via Dynamic Suppression of Language Priorscited by 3
  30. 2026CWCD: Category-Wise Contrastive Decoding for Structured Medical Report Generationcited by 3
  31. 2026A Practical Investigation of Training-free Relaxed Speculative Decodingcited by 3
  32. 2026HulluEdit: Single-Pass Evidence-Consistent Subspace Editing for Mitigating Hallucinations in Large Vision-Language Modelscited by 2
  33. 2026Mitigating Context-Memory Conflicts in LLMs through Dynamic Cognitive Reconciliation Decodingcited by 2
  34. 2026Reading the Finetuning Prior: Verbatim Content Recovery via Contrastive Decoding Diffingcited by 2
  35. 2024Chain-of-Thought Augmentation with Logit Contrast for Enhanced Reasoning in Language Modelscited by 1
  36. 2024Tuning Language Models by Mixture-of-Depths Ensemblecited by 1
  37. 2024What's New in My Data? Novelty Exploration via Contrastive Generationcited by 1
  38. 2024Dr. SoW: Density Ratio of Strong-over-weak LLMs for Reducing the Cost of Human Annotation in Preference Tuningcited by 1

50 most cited of 74 All 74 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2026DoubtLess: Training-Free Contrastive Decoding for Efficient Reasoning
  2. 2024Multilingual Contrastive Decoding via Language-Agnostic Layers Skipping
  3. 2023Are LLMs Rigorous Logical Reasoners? Empowering Natural Language Proof Generation by Stepwise Decoding with Contrastive Learning
  4. 2025Contrastive Decoding for Synthetic Data Generation in Low-Resource Language Modeling
  5. 2026Thinking Before Constraining: A Unified Decoding Framework for Large Language Models
  6. 2025Sample Smart, Not Hard: Correctness-First Decoding for Better Reasoning in LLMs
  7. 2025Robust Answers, Fragile Logic: Probing the Decoupling Hypothesis in LLM Reasoning
  8. 2025Language Models Can See Better: Visual Contrastive Decoding For LLM Multimodal Reasoning