mlsanityData to 7 Oct 2026

Stealing the Decoding Algorithms of Language Models

arXiv 2303.04729 ↗First posted 8 Mar 2023cs.LG cs.CL cs.CRcited by 19

Not on the map. Why?

Abstract (arXiv)

A key component of generating text from modern language models (LM) is the selection and tuning of decoding algorithms. These algorithms determine how to generate text from the internal probability distribution generated by the LM. The process of choosing a decoding algorithm and tuning its hyperparameters takes significant time, manual effort, and computation, and it also requires extensive human evaluation. Therefore, the identity and hyperparameters of such decoding algorithms are considered to be extremely valuable to their owners. In this work, we show, for the first time, that an adversary with typical API access to an LM can steal the type and hyperparameters of its decoding algorithms at very low monetary costs. Our attack is effective against popular LMs used in text generation APIs, including GPT-2, GPT-3 and GPT-Neo. We demonstrate the feasibility of stealing such information with only a few dollars, e.g., 0.8, 1, 4, and 40 for the four versions of GPT-3.

Key facts. Cited by 19 papers in the map (data through 7 Oct 2026). Builds on 44 of its 60 references that are in the map. First posted on arXiv 8 Mar 2023.

Builds on

44 of 60 references.

References · 44

  1. 2017Attention Is All You Needcited by 59,932
  2. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,204
  3. 2020Language Models are Few-Shot Learnerscited by 32,199
  4. 2019Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformercited by 14,491
  5. 2014Neural Machine Translation by Jointly Learning to Align and Translatecited by 4,530
  6. 2022OPT: Open Pre-trained Transformer Language Modelscited by 3,464
  7. 2019The Curious Case of Neural Text Degenerationcited by 2,881
  8. 2020The Pile: An 800GB Dataset of Diverse Text for Language Modelingcited by 2,346
  9. 2016Membership Inference Attacks against Machine Learning Modelscited by 2,221
  10. 2020Extracting Training Data from Large Language Modelscited by 2,144
  11. 2018Hierarchical Neural Story Generationcited by 1,182
  12. 2017BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chaincited by 1,052
32 more
  1. 2016Abstractive Text Summarization Using Sequence-to-Sequence RNNs and Beyondcited by 809
  2. 2016Stealing Machine Learning Models via Prediction APIscited by 602
  3. 2022Contrastive Decoding: Open-ended Text Generation as Optimizationcited by 598
  4. 2013Generating Sequences With Recurrent Neural Networkscited by 549
  5. 2021MAUVE: Measuring the Gap Between Neural Text and Human Text using Divergence Frontierscited by 466
  6. 2016Diverse Beam Search: Decoding Diverse Solutions from Neural Sequence Modelscited by 299
  7. 2018Knockoff Nets: Stealing Functionality of Black-Box Modelscited by 285
  8. 2020Evaluation of Text Generation: A Surveycited by 275
  9. 2020Extractive Summarization as Text Matchingcited by 189
  10. 2022Repairing the Cracked Foundation: A Survey of Obstacles in Evaluation Practices for Generated Textcited by 154
  11. 2017Controlling Linguistic Style Aspects in Neural Language Generationcited by 147
  12. 2019Thieves on Sesame Street! Model Extraction of BERT-based APIscited by 136
  13. 2019Unifying Human and Statistical Evaluation for Natural Language Generationcited by 127
  14. 2022Truncation Sampling as Language Model Desmoothingcited by 121
  15. 2018Learning to Write with Cooperative Discriminatorscited by 119
  16. 2018Stealing Hyperparameters in Machine Learningcited by 115
  17. 2018Sharp Nearby, Fuzzy Far Away: How Neural Language Models Use Contextcited by 113
  18. 2021Backdoor Attacks on Pre-trained Models by Layerwise Weight Poisoningcited by 103
  19. 2018Language GANs Falling Shortcited by 97
  20. 2019DAWN: Dynamic Adversarial Watermarking of Neural Networkscited by 94
  21. 2019Do Massively Pretrained Language Models Make Better Storytellers?cited by 90
  22. 2018Model Reconstruction from Model Explanationscited by 84
  23. 2020Trojaning Language Models for Fun and Profitcited by 80
  24. 2021Backdoor Pre-trained Models Can Transfer to Allcited by 79
  25. 2020Imitation Attacks and Defenses for Black-box Machine Translation Systemscited by 78
  26. 2021BadPre: Task-agnostic Backdoor Attacks to Pre-trained NLP Foundation Modelscited by 76
  27. 2022RankGen: Improving Text Generation with Large Ranking Modelscited by 72
  28. 2022Contrastive Search Is What You Need For Neural Text Generationcited by 58
  29. 2021Model Extraction and Adversarial Transferability, Your BERT is Vulnerable!cited by 51
  30. 2017Generating High-Quality and Informative Conversation Responses with Sequence-to-Sequence Modelscited by 42
  31. 2021Membership Inference on Word Embedding and Beyondcited by 34
  32. 2019Adversarial Neural Network Inversion via Auxiliary Knowledge Alignmentcited by 30

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 4; 2024: 9; 2025: 3; 2026: 3.

19 papers.

Sort:

By area:

  1. 2025Large Language Model Agent: A Survey on Methodology, Applications and Challengescited by 194
  2. 2024PLeak: Prompt Leaking Attacks against Large Language Model Applicationscited by 113
  3. 2023Privacy in Large Language Models: Attacks, Defenses and Future Directionscited by 48
  4. 2024No Free Lunch in LLM Watermarking: Trade-offs in Watermarking Design Choicescited by 47
  5. 2024SPML: A DSL for Defending Language Models Against Prompt Attackscited by 33
  6. 2024Large Model Based Agents: State-of-the-Art, Cooperation Paradigms, Security and Privacy, and Future Trendscited by 26
  7. 2024Transferable Embedding Inversion Attack: Uncovering Privacy Risks in Text Embeddings without Model Queriescited by 25
  8. 2024Black-Box Detection of Language Model Watermarkscited by 22
  9. 2023Tree Prompting: Efficient Task Adaptation without Fine-Tuningcited by 17
  10. 2024Regression-aware Inference with LLMscited by 13
  11. 2024Model Stealing for Any Low-Rank Language Modelcited by 9
  12. 2023Continual Learning From a Stream of APIscited by 6
7 more
  1. 2024On Evaluating The Performance of Watermarked Machine-Generated Texts Under Adversarial Attackscited by 4
  2. 2023Reverse-Engineering Decoding Strategies Given Blackbox Access to a Language Generation Systemcited by 3
  3. 2025Breaking Distortion-free Watermarks in Large Language Modelscited by 0
  4. 2025Relative-Based Scaling Law for Neural Language Modelscited by 0
  5. 2026DiMEx: Breaking the Cold Start Barrier in Data-Free Model Extraction via Latent Diffusion Priorscited by 0
  6. 2026From Instruction to Output: The Role of Prompting in Modern NLGcited by 0
  7. 2026Beyond Tokens: A Survey on Decoding Methods for Large Language and Vision-Language Modelscited by 0