mlsanityData to 7 Oct 2026

DExperts: Decoding-Time Controlled Text Generation with Experts and Anti-Experts

arXiv 2105.03023 ↗First posted 7 May 2021cs.CLcited by 434

Not on the map. Why?

Cited by 49 in LLM alignment / RLHF / preference optimisation, but placed by its own abstract, which didn’t fit.

Abstract (arXiv)

Despite recent advances in natural language generation, it remains challenging to control attributes of generated text. We propose DExperts: Decoding-time Experts, a decoding-time method for controlled text generation that combines a pretrained language model with "expert" LMs and/or "anti-expert" LMs in a product of experts. Intuitively, under the ensemble, tokens only get high probability if they are considered likely by the experts, and unlikely by the anti-experts. We apply DExperts to language detoxification and sentiment-controlled generation, where we outperform existing controllable generation methods on both automatic and human evaluations. Moreover, because DExperts operates only on the output of the pretrained LM, it is effective with (anti-)experts of smaller size, including when operating on GPT-3. Our work highlights the promise of tuning small LMs on text with (un)desirable attributes for efficient decoding-time steering.

Key facts. Cited by 434 papers in the map (data through 7 Oct 2026). Builds on 29 of its 53 references that are in the map. First posted on arXiv 7 May 2021.

Builds on

29 of 53 references.

References · 29

  1. 2020Language Models are Few-Shot Learnerscited by 32,199
  2. 2019BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehensioncited by 5,874
  3. 2021Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsitycited by 3,014
  4. 2019The Curious Case of Neural Text Degenerationcited by 2,881
  5. 2020Don't Stop Pretraining: Adapt Language Models to Domains and Taskscited by 1,548
  6. 2020RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Modelscited by 1,235
  7. 2018Hierarchical Neural Story Generationcited by 1,182
  8. 2015A Diversity-Promoting Objective Function for Neural Conversation Modelscited by 1,068
  9. 2019Plug and Play Language Models: A Simple Approach to Controlled Text Generationcited by 793
  10. 2019CTRL: A Conditional Transformer Language Model for Controllable Generationcited by 754
  11. 2015Image-based Recommendations on Styles and Substitutescited by 702
  12. 2019The Woman Worked as a Babysitter: On Biases in Language Generationcited by 540
17 more
  1. 2019Neural Text Generation with Unlikelihood Trainingcited by 489
  2. 2019Social Bias Frames: Reasoning about Social and Power Implications of Languagecited by 377
  3. 2020GeDi: Generative Discriminator Guided Sequence Generationcited by 377
  4. 2021FUDGE: Controlled Text Generation With Future Discriminatorscited by 374
  5. 2019Measurement and Fairnesscited by 299
  6. 2019Racial Bias in Hate Speech and Abusive Language Detection Datasetscited by 208
  7. 2018Delete, Retrieve, Generate: A Simple Approach to Sentiment and Style Transfercited by 191
  8. 2020Social Biases in NLP Models as Barriers for Persons with Disabilitiescited by 182
  9. 2017Controlling Linguistic Style Aspects in Neural Language Generationcited by 147
  10. 2019What makes a good conversation? How controllable attributes affect human judgmentscited by 140
  11. 2018Learning to Write with Cooperative Discriminatorscited by 119
  12. 2020Towards Controllable Biases in Language Generationcited by 112
  13. 2019Transforming Delete, Retrieve, Generate Approach for Controlled Text Style Transfercited by 81
  14. 2020The Radicalization Risks of GPT-3 and Advanced Neural Language Modelscited by 76
  15. 2020Exploring Controllable Text Generation Techniquescited by 55
  16. 2017Affect-LM: A Neural Language Model for Customizable Affective Text Generationcited by 50
  17. 2020Plug-and-Play Conversational Modelscited by 30

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 12; 2022: 56; 2023: 82; 2024: 121; 2025: 74; 2026: 89.

434 papers.

Sort:

By area:

  1. 2022Editing Models with Task Arithmeticcited by 1,076
  2. 2022Diffusion-LM Improves Controllable Text Generationcited by 1,006
  3. 2023CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquingcited by 751
  4. 2022Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learnedcited by 702
  5. 2023Mitigating Object Hallucinations in Large Vision-Language Models through Visual Contrastive Decodingcited by 671
  6. 2023Bias and Fairness in Large Language Models: A Surveycited by 626
  7. 2022Contrastive Decoding: Open-ended Text Generation as Optimizationcited by 598
  8. 2022ToxiGen: A Large-Scale Machine-Generated Dataset for Adversarial and Implicit Hate Speech Detectioncited by 419
  9. 2023Fine-Grained Human Feedback Gives Better Rewards for Language Model Trainingcited by 399
  10. 2022RLPrompt: Optimizing Discrete Text Prompts with Reinforcement Learningcited by 361
  11. 2023Trusting Your Evidence: Hallucinate Less with Context-aware Decodingcited by 319
  12. 2022Generating Sequences by Learning to Self-Correctcited by 271
38 more
  1. 2022Quark: Controllable Text Generation with Reinforced Unlearningcited by 236
  2. 2022Extracting Latent Steering Vectors from Pretrained Language Modelscited by 217
  3. 2021Societal Biases in Language Generation: Progress and Challengescited by 201
  4. 2022Generating Training Data with Language Models: Towards Zero-Shot Language Understandingcited by 199
  5. 2023"Kelly is a Warm Person, Joseph is a Role Model": Gender Biases in LLM-Generated Reference Letterscited by 196
  6. 2022SSD-LM: Semi-autoregressive Simplex-based Diffusion Language Model for Text Generation and Modular Controlcited by 191
  7. 2023Automatically Auditing Large Language Models via Discrete Optimizationcited by 187
  8. 2024A Roadmap to Pluralistic Alignmentcited by 175
  9. 2024COLD-Attack: Jailbreaking LLMs with Stealthiness and Controllabilitycited by 162
  10. 2024From Decoding to Meta-Generation: Inference-time Algorithms for Large Language Modelscited by 136
  11. 2021DEMix Layers: Disentangling Domains for Modular Language Modelingcited by 130
  12. 2022A Survey of Controllable Text Generation using Transformer-based Pre-trained Language Modelscited by 128
  13. 2024A Thorough Examination of Decoding Methods in the Era of LLMscited by 112
  14. 2024Weak-to-Strong Jailbreaking on Large Language Modelscited by 100
  15. 2024ARGS: Alignment as Reward-Guided Searchcited by 99
  16. 2024BetterV: Controlled Verilog Generation with Discriminative Guidancecited by 99
  17. 2022Prompt Compression and Contrastive Conditioning for Controllability and Toxicity Reduction in Language Modelscited by 96
  18. 2023Guiding Large Language Models via Directional Stimulus Promptingcited by 94
  19. 2023Amortizing intractable inference in large language modelscited by 94
  20. 2024Steering Language Model Refusal with Sparse Autoencoderscited by 89
  21. 2024Tuning Language Models by Proxycited by 88
  22. 2024Probabilistic Inference in Language Models via Twisted Sequential Monte Carlocited by 85
  23. 2024Reversing the Forget-Retain Objectives: An Efficient LLM Unlearning Framework from Logit Differencecited by 85
  24. 2024Aligning to Thousands of Preferences via System Message Generalizationcited by 84
  25. 2022Language Generation Models Can Cause Harm: So What Can We Do About It? An Actionable Surveycited by 80
  26. 2023Reward-Augmented Decoding: Efficient Controlled Text Generation With a Unidirectional Reward Modelcited by 79
  27. 2023Alleviating Hallucinations of Large Language Models through Induced Hallucinationscited by 78
  28. 2021On the Safety of Conversational Models: Taxonomy, Dataset, and Benchmarkcited by 73
  29. 2023Contrastive Decoding Improves Reasoning in Large Language Modelscited by 71
  30. 2021Controlled Text Generation as Continuous Optimization with Multiple Constraintscited by 70
  31. 2023Word Embeddings Are Steers for Language Modelscited by 70
  32. 2022Exploring the Limits of Domain-Adaptive Training for Detoxifying Large-Scale Language Modelscited by 67
  33. 2022ProGen: Progressive Zero-shot Dataset Generation via In-context Feedbackcited by 67
  34. 2022Mix and Match: Learning-free Controllable Text Generation using Energy Language Modelscited by 66
  35. 2022Prompt Consistency for Zero-Shot Task Generalizationcited by 66
  36. 2022Gradient-Based Constrained Sampling from Language Modelscited by 64
  37. 2022Keep Me Updated! Memory Management in Long-term Conversationscited by 62
  38. 2024Interpretable Contrastive Monte Carlo Tree Search Reasoningcited by 62

50 most cited of 434 All 434 (JSON)