mlsanityData to 7 Oct 2026

GeDi: Generative Discriminator Guided Sequence Generation

arXiv 2009.06367 ↗First posted 14 Sep 2020cs.CL cs.LGcited by 377

Not on the map. Why?

Cited by 31 in LLM alignment / RLHF / preference optimisation, but placed by its own abstract, which didn’t fit.

Abstract (arXiv)

While large-scale language models (LMs) are able to imitate the distribution of natural language well enough to generate realistic text, it is difficult to control which regions of the distribution they generate. This is especially problematic because datasets used for training large LMs usually contain significant toxicity, hate, bias, and negativity. We propose GeDi as an efficient method for using smaller LMs as generative discriminators to guide generation from large LMs to make them safer and more controllable. GeDi guides generation at each step by computing classification probabilities for all possible next tokens via Bayes rule by normalizing over two class-conditional distributions; one conditioned on the desired attribute, or control code, and another conditioned on the undesired attribute, or anti control code. We find that GeDi gives stronger controllability than the state of the art method while also achieving generation speeds more than 30 times faster. Additionally, training GeDi on only four topics allows us to controllably generate new topics zero-shot from just a keyword, unlocking a new capability that previous controllable generation methods do not have. Lastly, we show that GeDi can make GPT-2 (1.5B parameters) significantly less toxic without sacrificing linguistic quality, making it by far the most practical existing method for detoxifying large language models while maintaining a fast generation speed.

Key facts. Cited by 377 papers in the map (data through 7 Oct 2026). First posted on arXiv 14 Sep 2020.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 31; 2022: 78; 2023: 85; 2024: 80; 2025: 45; 2026: 58.

377 papers.

Sort:

By area:

  1. 2022Training language models to follow instructions with human feedbackcited by 15,712
  2. 2021Prefix-Tuning: Optimizing Continuous Prompts for Generationcited by 3,878
  3. 2022OPT: Open Pre-trained Transformer Language Modelscited by 3,464
  4. 2021Decision Transformer: Reinforcement Learning via Sequence Modelingcited by 1,528
  5. 2023Inference-Time Intervention: Eliciting Truthful Answers from a Language Modelcited by 1,259
  6. 2022Diffusion Models: A Comprehensive Survey of Methods and Applicationscited by 1,107
  7. 2022Diffusion-LM Improves Controllable Text Generationcited by 1,006
  8. 2023CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquingcited by 751
  9. 2023Bias and Fairness in Large Language Models: A Surveycited by 626
  10. 2021Prompt Programming for Large Language Models: Beyond the Few-Shot Paradigmcited by 599
  11. 2022Using DeepSpeed and Megatron to Train Megatron-Turing NLG 530B, A Large-Scale Generative Language Modelcited by 557
  12. 2023RRHF: Rank Responses to Align Language Models with Human Feedback without tearscited by 454
38 more
  1. 2021DExperts: Decoding-Time Controlled Text Generation with Experts and Anti-Expertscited by 434
  2. 2022ToxiGen: A Large-Scale Machine-Generated Dataset for Adversarial and Implicit Hate Speech Detectioncited by 419
  3. 2022Emergent World Representations: Exploring a Sequence Model Trained on a Synthetic Taskcited by 417
  4. 2023Fine-Grained Human Feedback Gives Better Rewards for Language Model Trainingcited by 399
  5. 2021FUDGE: Controlled Text Generation With Future Discriminatorscited by 374
  6. 2022CodeRL: Mastering Code Generation through Pretrained Models and Deep Reinforcement Learningcited by 342
  7. 2021Self-Diagnosis and Self-Debiasing: A Proposal for Reducing Corpus-Based Bias in NLPcited by 336
  8. 2023CodeT5+: Open Code Large Language Models for Code Understanding and Generationcited by 327
  9. 2021Unsolved Problems in ML Safetycited by 291
  10. 2022Generating Sequences by Learning to Self-Correctcited by 271
  11. 2024Model Merging in LLMs, MLLMs, and Beyond: Methods, Theories, Applications and Opportunitiescited by 247
  12. 2022Quark: Controllable Text Generation with Reinforced Unlearningcited by 236
  13. 2022Extracting Latent Steering Vectors from Pretrained Language Modelscited by 217
  14. 2022Multi-Game Decision Transformerscited by 212
  15. 2021Societal Biases in Language Generation: Progress and Challengescited by 201
  16. 2022Generating Training Data with Language Models: Towards Zero-Shot Language Understandingcited by 199
  17. 2022Re3: Generating Longer Stories With Recursive Reprompting and Revisioncited by 193
  18. 2022SSD-LM: Semi-autoregressive Simplex-based Diffusion Language Model for Text Generation and Modular Controlcited by 191
  19. 2023Automatically Auditing Large Language Models via Discrete Optimizationcited by 187
  20. 2021Challenges in Detoxifying Language Modelscited by 169
  21. 2024COLD-Attack: Jailbreaking LLMs with Stealthiness and Controllabilitycited by 162
  22. 2022COLD Decoding: Energy-based Constrained Text Generation with Langevin Dynamicscited by 159
  23. 2024A Comprehensive Study of Knowledge Editing for Large Language Modelscited by 157
  24. 2023Large Language Models for Code: Security Hardening and Adversarial Testingcited by 149
  25. 2023Likelihood-Based Diffusion Language Modelscited by 147
  26. 2022Offline RL for Natural Language Generation with Implicit Language Q Learningcited by 137
  27. 2023Controlled Decoding from Language Modelscited by 133
  28. 2022A Survey of Controllable Text Generation using Transformer-based Pre-trained Language Modelscited by 128
  29. 2021Detoxifying Language Models Risks Marginalizing Minority Voicescited by 106
  30. 2024ARGS: Alignment as Reward-Guided Searchcited by 99
  31. 2024BetterV: Controlled Verilog Generation with Discriminative Guidancecited by 99
  32. 2022Prompt Compression and Contrastive Conditioning for Controllability and Toxicity Reduction in Language Modelscited by 96
  33. 2023Guiding Large Language Models via Directional Stimulus Promptingcited by 94
  34. 2023Stay on topic with Classifier-Free Guidancecited by 90
  35. 2024Tuning Language Models by Proxycited by 88
  36. 2024Probabilistic Inference in Language Models via Twisted Sequential Monte Carlocited by 85
  37. 2024Aligning to Thousands of Preferences via System Message Generalizationcited by 84
  38. 2022Language Generation Models Can Cause Harm: So What Can We Do About It? An Actionable Surveycited by 80

50 most cited of 377 All 377 (JSON)