mlsanityData to 7 Oct 2026

COLD Decoding: Energy-based Constrained Text Generation with Langevin Dynamics

arXiv 2202.11705 ↗First posted 23 Feb 2022cs.CL cs.AI cs.LGcited by 159

Not on the map. Why?

Cited by 21 in LLM safety and security, but placed by its own abstract, which didn’t fit.

Abstract (arXiv)

Many applications of text generation require incorporating different constraints to control the semantics or style of generated text. These constraints can be hard (e.g., ensuring certain keywords are included in the output) and soft (e.g., contextualizing the output with the left- or right-hand context). In this paper, we present Energy-based Constrained Decoding with Langevin Dynamics (COLD), a decoding framework which unifies constrained generation as specifying constraints through an energy function, then performing efficient differentiable reasoning over the constraints through gradient-based sampling. COLD decoding is a flexible framework that can be applied directly to off-the-shelf left-to-right language models without the need for any task-specific fine-tuning, as demonstrated through three challenging text generation applications: lexically-constrained generation, abductive reasoning, and counterfactual reasoning. Our experiments on these constrained generation tasks point to the effectiveness of our approach, both in terms of automatic and human evaluation.

Key facts. Cited by 159 papers in the map (data through 7 Oct 2026). Builds on 37 of its 75 references that are in the map. First posted on arXiv 23 Feb 2022.

Builds on

37 of 75 references.

References · 37

  1. 2017Attention Is All You Needcited by 59,932
  2. 2020Language Models are Few-Shot Learnerscited by 32,199
  3. 2021Evaluating Large Language Models Trained on Codecited by 8,053
  4. 2019BERTScore: Evaluating Text Generation with BERTcited by 5,462
  5. 2019Generative Modeling by Estimating Gradients of the Data Distributioncited by 4,067
  6. 2019The Curious Case of Neural Text Degenerationcited by 2,881
  7. 2014CIDEr: Consensus-based Image Description Evaluationcited by 2,027
  8. 2018Hierarchical Neural Story Generationcited by 1,182
  9. 2019Plug and Play Language Models: A Simple Approach to Controlled Text Generationcited by 793
  10. 2020GeDi: Generative Discriminator Guided Sequence Generationcited by 377
  11. 2021FUDGE: Controlled Text Generation With Future Discriminatorscited by 374
  12. 2019Abductive Commonsense Reasoningcited by 300
25 more
  1. 2017Toward Controlled Generation of Textcited by 271
  2. 2017Lexically Constrained Decoding for Sequence Generation Using Grid Beam Searchcited by 234
  3. 2020NeuroLogic Decoding: (Un)supervised Neural Text Generation with Predicate Logic Constraintscited by 146
  4. 2020Enabling Language Models to Fill in the Blankscited by 121
  5. 2020A Distributional Approach to Controlled Text Generationcited by 117
  6. 2019On the Anatomy of MCMC-Based Maximum Likelihood Learning of Energy-Based Modelscited by 105
  7. 2020Learning Latent Space Energy-Based Prior Modelcited by 104
  8. 2020Residual Energy-Based Models for Text Generationcited by 95
  9. 2018CGMH: Constrained Sentence Generation by Metropolis-Hastings Samplingcited by 93
  10. 2017Generating Sentences by Editing Prototypescited by 88
  11. 2019Counterfactual Story Reasoning and Generationcited by 87
  12. 2021Controlled Text Generation as Continuous Optimization with Multiple Constraintscited by 70
  13. 2018Sampling Can Be Faster Than Optimizationcited by 68
  14. 2020Back to the Future: Unsupervised Backprop-based Decoding for Counterfactual and Abductive Commonsense Reasoningcited by 65
  15. 2020No MCMC for me: Amortized sampling for fast and stable training of energy-based modelscited by 64
  16. 2020Generative PointNet: Deep Energy-Based Learning on Unordered Point Sets for 3D Generation, Reconstruction and Classificationcited by 57
  17. 2021Exposing the Implicit Energy Networks behind Masked Language Models via Metropolis--Hastingscited by 37
  18. 2019Text Infillingcited by 32
  19. 2018Deep Generative Models with Learnable Knowledge Constraintscited by 23
  20. 2021Parallel and Flexible Sampling from Autoregressive Models via Langevin Dynamicscited by 19
  21. 2021Don't Take It Literally: An Edit-Invariant Sequence Loss for Text Generationcited by 17
  22. 2017Towards Decoding as Continuous Optimization in Neural Machine Translationcited by 15
  23. 2019Global Autoregressive Models for Data-Efficient Sequence Learningcited by 15
  24. 2020Directed Beam Search: Plug-and-Play Lexically Constrained Language Generationcited by 12
  25. 2020Reflective Decoding: Beyond Unidirectional Generation with Off-the-Shelf Language Modelscited by 9

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2022: 18; 2023: 32; 2024: 51; 2025: 35; 2026: 23.

159 papers.

Sort:

By area:

  1. 2023Universal and Transferable Adversarial Attacks on Aligned Language Modelscited by 2,914
  2. 2022Diffusion-LM Improves Controllable Text Generationcited by 1,006
  3. 2022RLPrompt: Optimizing Discrete Text Prompts with Reinforcement Learningcited by 361
  4. 2022Generating Sequences by Learning to Self-Correctcited by 271
  5. 2022SSD-LM: Semi-autoregressive Simplex-based Diffusion Language Model for Text Generation and Modular Controlcited by 191
  6. 2023Automatically Auditing Large Language Models via Discrete Optimizationcited by 187
  7. 2024A Roadmap to Pluralistic Alignmentcited by 175
  8. 2024COLD-Attack: Jailbreaking LLMs with Stealthiness and Controllabilitycited by 162
  9. 2023Controlled Decoding from Language Modelscited by 133
  10. 2024Derivative-Free Guidance in Continuous and Discrete Diffusion Models with Soft Value-Based Decodingcited by 127
  11. 2024BoNBoN Alignment for Large Language Models and the Sweetness of Best-of-n Samplingcited by 120
  12. 2023Aligning Language Models with Preferences through f-divergence Minimizationcited by 118
38 more
  1. 2023A Critical Evaluation of Evaluations for Long-form Question Answeringcited by 111
  2. 2024Decoding-Time Language Model Alignment with Multiple Objectivescited by 90
  3. 2023Controlled Text Generation with Natural Language Instructionscited by 87
  4. 2024Probabilistic Inference in Language Models via Twisted Sequential Monte Carlocited by 85
  5. 2023Sequential Monte Carlo Steering of Large Language Models using Probabilistic Programscited by 79
  6. 2023Evaluating Large Language Models on Controlled Generation Taskscited by 73
  7. 2024FlipAttack: Jailbreak LLMs via Flippingcited by 73
  8. 2022RankGen: Improving Text Generation with Large Ranking Modelscited by 72
  9. 2022NaturalProver: Grounded Mathematical Proof Generation with Language Modelscited by 72
  10. 2024Imprompter: Tricking LLM Agents into Improper Tool Usecited by 66
  11. 2024RigorLLM: Resilient Guardrails for Large Language Models against Undesired Contentcited by 65
  12. 2022Gradient-Based Constrained Sampling from Language Modelscited by 64
  13. 2024DeAL: Decoding-time Alignment for Large Language Modelscited by 55
  14. 2024Grammar-Aligned Decodingcited by 54
  15. 2024Don't Say No: Jailbreaking LLM by Suppressing Refusalcited by 53
  16. 2025Edit Flows: Flow Matching with Edit Operationscited by 53
  17. 2025Syntactic and Semantic Control of Large Language Models via Sequential Monte Carlocited by 45
  18. 2022LayoutFormer++: Conditional Graphic Layout Generation via Constraint Serialization and Decoding Space Restrictioncited by 43
  19. 2023KCTS: Knowledge-Constrained Tree Search Decoding with Token-Level Hallucination Detectioncited by 43
  20. 2024Space Group Informed Transformer for Crystalline Materials Generationcited by 40
  21. 2024Formal-LLM: Integrating Formal Language and Natural Language for Controllable LLM-based Agentscited by 39
  22. 2025Dynamic Search for Inference-Time Alignment in Diffusion Modelscited by 36
  23. 2024Constrained Decoding for Secure Code Generationcited by 35
  24. 2024Automatic Pseudo-Harmful Prompt Generation for Evaluating False Refusals in Large Language Modelscited by 35
  25. 2022A Distributional Lens for Multi-Aspect Controllable Text Generationcited by 33
  26. 2023Opportunities and Challenges in Neural Dialog Tutoringcited by 33
  27. 2023Inference-Time Policy Adapters (IPA): Tailoring Extreme-Scale LMs without Fine-tuningcited by 33
  28. 2024DeepEdit: Knowledge Editing as Decoding with Constraintscited by 33
  29. 2024Adaptable Logical Control for Large Language Modelscited by 30
  30. 2022Composable Text Controls in Latent Space with ODEscited by 28
  31. 2022Controllable Text Generation via Probability Density Estimation in the Latent Spacecited by 27
  32. 2023Denevil: Towards Deciphering and Navigating the Ethical Values of Large Language Models via Instruction Learningcited by 26
  33. 2024Benchmarking Large Language Models on Controllable Generation under Diversified Instructionscited by 26
  34. 2025Riddle Me This! Stealthy Membership Inference for Retrieval-Augmented Generationcited by 26
  35. 2023BOLT: Fast Energy-based Controlled Text Generation with Tunable Biasescited by 24
  36. 2025Antidistillation Samplingcited by 24
  37. 2022CORE: A Retrieve-then-Edit Framework for Counterfactual Data Generationcited by 23
  38. 2024SafeInfer: Context Adaptive Decoding Time Safety Alignment for Large Language Modelscited by 22

50 most cited of 159 All 159 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2024Improving Open-Ended Text Generation via Adaptive Decoding
  2. 2026Thinking Before Constraining: A Unified Decoding Framework for Large Language Models
  3. 2025A Survey on Parallel Text Generation: From Parallel Decoding to Diffusion Language Models
  4. 2024Guiding LLMs The Right Way: Fast, Non-Invasive Constrained Generation
  5. 2024A Comprehensive Survey of Accelerated Generation Techniques in Large Language Models
  6. 2025MAGNET: Augmenting Generative Decoders with Representation Learning and Infilling Capabilities
  7. 2022ELMER: A Non-Autoregressive Pre-trained Language Model for Efficient and Effective Text Generation
  8. 2026Unified Energy for Invariant and Independent Decoding in Diffusion Language Models