mlsanityData to 14 Oct 2026

Controllable Natural Language Generation with Contrastive Prefixes

arXiv 2202.13257 ↗First posted 27 Feb 2022cs.CLcited by 75

Not on the map. Why?

Abstract (arXiv)

To guide the generation of large pretrained language models (LM), previous work has focused on directly fine-tuning the language model or utilizing an attribute discriminator. In this work, we propose a novel lightweight framework for controllable GPT2 generation, which utilizes a set of small attribute-specific vectors, called prefixes, to steer natural language generation. Different from prefix-tuning, where each prefix is trained independently, we take the relationship among prefixes into consideration and train multiple prefixes simultaneously. We propose a novel supervised method and also an unsupervised method to train the prefixes for single-aspect control while the combination of these two methods can achieve multi-aspect control. Experimental results on both single-aspect and multi-aspect control show that our methods can guide generation towards the desired attributes while keeping high linguistic quality.

Key facts. Cited by 75 papers in the map (data through 14 Oct 2026). Builds on 17 of its 23 references that are in the map. First posted on arXiv 27 Feb 2022.

Builds on

17 of 23 references.

References · 17

  1. 2019RoBERTa: A Robustly Optimized BERT Pretraining Approachcited by 13,372
  2. 2017Neural Discrete Representation Learningcited by 4,712
  3. 2021Zero-Shot Text-to-Image Generationcited by 4,288
  4. 2021Prefix-Tuning: Optimizing Continuous Prompts for Generationcited by 3,915
  5. 2016Categorical Reparameterization with Gumbel-Softmaxcited by 2,739
  6. 2015Character-level Convolutional Networks for Text Classificationcited by 2,330
  7. 2020RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Modelscited by 1,255
  8. 2016The Concrete Distribution: A Continuous Relaxation of Discrete Random Variablescited by 1,148
  9. 2019Plug and Play Language Models: A Simple Approach to Controlled Text Generationcited by 799
  10. 2019CTRL: A Conditional Transformer Language Model for Controllable Generationcited by 762
  11. 2020GeDi: Generative Discriminator Guided Sequence Generationcited by 378
  12. 2021FUDGE: Controlled Text Generation With Future Discriminatorscited by 376
5 more
  1. 2017Toward Controlled Generation of Textcited by 272
  2. 2021Challenges in Detoxifying Language Modelscited by 169
  3. 2017Controlling Linguistic Style Aspects in Neural Language Generationcited by 149
  4. 2018Learning to Write with Cooperative Discriminatorscited by 119
  5. 2021Attribute Alignment: Controlling Text Generation from Pre-trained Language Modelscited by 27

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2022: 15; 2023: 23; 2024: 25; 2025: 8; 2026: 4.

75 papers.

Sort:

By area:

  1. 2023Instruction Tuning for Large Language Models: A Surveycited by 623
  2. 2022RLPrompt: Optimizing Discrete Text Prompts with Reinforcement Learningcited by 366
  3. 2024A Survey on Recent Advances in LLM-Based Multi-turn Dialogue Systemscited by 170
  4. 2023Large Language Models for Code: Security Hardening and Adversarial Testingcited by 155
  5. 2022Gradient-Based Constrained Sampling from Language Modelscited by 64
  6. 2022Generative Knowledge Graph Construction: A Reviewcited by 49
  7. 2024Confidence Matters: Revisiting Intrinsic Self-Correction Capabilities of Large Language Modelscited by 47
  8. 2022Critic-Guided Decoding for Controlled Text Generationcited by 45
  9. 2022A Distributional Lens for Multi-Aspect Controllable Text Generationcited by 33
  10. 2022Unified Detoxifying and Debiasing in Language Generation via Inference-time Adaptive Optimizationcited by 33
  11. 2022DisCup: Discriminator Cooperative Unlikelihood Prompt-tuning for Controllable Text Generationcited by 30
  12. 2023Trends in Integration of Knowledge and Large Language Models: A Survey and Taxonomy of Methods, Benchmarks, and Applicationscited by 28
38 more
  1. 2022Controllable Text Generation via Probability Density Estimation in the Latent Spacecited by 27
  2. 2024Benchmarking Large Language Models on Controllable Generation under Diversified Instructionscited by 27
  3. 2023Mixture of Soft Prompts for Controllable Data Generationcited by 23
  4. 2022Tailor: A Prompt-Based Approach to Attribute-Based Controlled Text Generationcited by 22
  5. 2022PromptEHR: Conditional Electronic Healthcare Records Generation with Prompt Learningcited by 22
  6. 2024Personalized Text Generation with Fine-Grained Linguistic Controlcited by 20
  7. 2023From Google Gemini to OpenAI Q* (Q-Star): A Survey of Reshaping the Generative Artificial Intelligence (AI) Research Landscapecited by 18
  8. 2024Contrastive Instruction Tuningcited by 17
  9. 2024Joint Visual and Text Prompting for Improved Object-Centric Perception with Multimodal Large Language Modelscited by 17
  10. 2024Walking in Others' Shoes: How Perspective-Taking Guides Large Language Models in Reducing Toxicity and Biascited by 17
  11. 2024Controlled Text Generation for Large Language Model with Dynamic Attribute Graphscited by 16
  12. 2024Reinforcement Learning with Token-level Feedback for Controllable Text Generationcited by 16
  13. 2024LLM-based multi-agent poetry generation in non-cooperative environmentscited by 15
  14. 2024Dynamic Multi-Reward Weighting for Multi-Style Controllable Generationcited by 13
  15. 2023MacLaSa: Multi-Aspect Controllable Text Generation via Efficient Sampling from Compact Latent Spacecited by 12
  16. 2023SimOAP: Improve Coherence and Consistency in Persona-based Dialogue Generation via Over-sampling and Post-evaluationcited by 11
  17. 2022An Extensible Plug-and-Play Method for Multi-Aspect Controllable Text Generationcited by 10
  18. 2023Air-Decoding: Attribute Distribution Reconstruction for Decoding-Time Controllable Text Generationcited by 10
  19. 2023Focused Prefix Tuning for Controllable Text Generationcited by 9
  20. 2023Unpacking the Ethical Value Alignment in Big Modelscited by 9
  21. 2022FAST: Improving Controllability for Text Generation with Feedback Aware Self-Trainingcited by 7
  22. 2023Seen to Unseen: Exploring Compositional Generalization of Multi-Attribute Controllable Dialogue Generationcited by 7
  23. 2023Dialogue for Prompting: a Policy-Gradient-Based Discrete Prompt Generation for Few-shot Learningcited by 7
  24. 2024Controllable Text Generation in the Instruction-Tuning Eracited by 7
  25. 2024Concentrate Attention: Towards Domain-Generalizable Prompt Optimization for Language Modelscited by 7
  26. 2024Attribute Controlled Fine-tuning for Large Language Models: A Case Study on Detoxificationcited by 7
  27. 2024FreeCtrl: Constructing Control Centers with Feedforward Layers for Learning-Free Controllable Text Generationcited by 6
  28. 2022DuNST: Dual Noisy Self Training for Semi-Supervised Controllable Text Generationcited by 5
  29. 2024Control Large Language Models via Divide and Conquercited by 5
  30. 2025Scaling Towards the Information Boundary of Instruction Sets: The Infinity Instruct Subject Technical Reportcited by 5
  31. 2022Zero-Label Prompt Selectioncited by 4
  32. 2023Pre-training Multi-party Dialogue Models with Latent Discourse Inferencecited by 4
  33. 2023Attribute Controlled Dialogue Promptingcited by 3
  34. 2023Controllable Text Generation with Residual Memory Transformercited by 3
  35. 2024LiFi: Lightweight Controlled Text Generation with Fine-Grained Control Codescited by 3
  36. 2024Continuous Language Model Interpolation for Dynamic and Controllable Text Generationcited by 3
  37. 2024Multi-Aspect Controllable Text Generation with Disentangled Counterfactual Augmentationcited by 3
  38. 2024ASTPrompter: Preference-Aligned Automated Language Model Red-Teaming to Generate Low-Perplexity Unsafe Promptscited by 3

50 most cited of 75 All 75 (JSON)