mlsanityData to 7 Oct 2026

Learning How to Ask: Querying LMs with Mixtures of Soft Prompts

arXiv 2104.06599 ↗First posted 14 Apr 2021cs.CL cs.LGcited by 376

Not on the map. Why?

Abstract (arXiv)

Natural-language prompts have recently been used to coax pretrained language models into performing other AI tasks, using a fill-in-the-blank paradigm (Petroni et al., 2019) or a few-shot extrapolation paradigm (Brown et al., 2020). For example, language models retain factual knowledge from their training corpora that can be extracted by asking them to "fill in the blank" in a sentential prompt. However, where does this prompt come from? We explore the idea of learning prompts by gradient descent -- either fine-tuning prompts taken from previous work, or starting from random initialization. Our prompts consist of "soft words," i.e., continuous vectors that are not necessarily word type embeddings from the language model. Furthermore, for each task, we optimize a mixture of prompts, learning which prompts are most effective and how to ensemble them. Across multiple English LMs and tasks, our approach hugely outperforms previous methods, showing that the implicit factual knowledge in language models was previously underestimated. Moreover, this knowledge is cheap to elicit: random initialization is nearly as good as informed initialization.

Key facts. Cited by 376 papers in the map (data through 7 Oct 2026). Builds on 21 of its 39 references that are in the map. First posted on arXiv 14 Apr 2021.

Builds on

21 of 39 references.

References · 21

  1. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,204
  2. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  3. 2020Language Models are Few-Shot Learnerscited by 32,199
  4. 2019RoBERTa: A Robustly Optimized BERT Pretraining Approachcited by 13,266
  5. 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,160
  6. 2019BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehensioncited by 5,874
  7. 2021Prefix-Tuning: Optimizing Continuous Prompts for Generationcited by 3,878
  8. 2019PIQA: Reasoning about Physical Commonsense in Natural Languagecited by 2,818
  9. 2018Deep contextualized word representationscited by 2,285
  10. 2019Language Models as Knowledge Bases?cited by 1,922
  11. 2016ConceptNet 5.5: An Open Multilingual Graph of General Knowledgecited by 1,065
  12. 2019How Can We Know What Language Models Know?cited by 953
9 more
  1. 2020Exploiting Cloze Questions for Few Shot Text Classification and Natural Language Inferencecited by 934
  2. 2020A Primer in BERTology: What we know about how BERT workscited by 807
  3. 2020It's Not Just Size That Matters: Small Language Models Are Also Few-Shot Learnerscited by 668
  4. 2016Neural Architectures for Named Entity Recognitioncited by 548
  5. 2020AutoPrompt: Eliciting Knowledge from Language Models with Automatically Generated Promptscited by 246
  6. 2019oLMpics -- On what Language Model Pre-training Capturescited by 161
  7. 2021BERTese: Learning to Speak to BERTcited by 82
  8. 2019Inducing Relational Knowledge from BERTcited by 75
  9. 2019LINSPECTOR WEB: A Multilingual Probing Suite for Word Representationscited by 5

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 35; 2022: 98; 2023: 132; 2024: 67; 2025: 33; 2026: 11.

376 papers.

Sort:

By area:

  1. 2021The Power of Scale for Parameter-Efficient Prompt Tuningcited by 3,769
  2. 2021Finetuned Language Models Are Zero-Shot Learnerscited by 3,747
  3. 2023QLoRA: Efficient Finetuning of Quantized LLMscited by 3,355
  4. 2021Pre-train, Prompt, and Predict: A Systematic Survey of Prompting Methods in Natural Language Processingcited by 2,222
  5. 2022Large Language Models Are Human-Level Prompt Engineerscited by 1,020
  6. 2022Few-Shot Parameter-Efficient Fine-Tuning is Better and Cheaper than In-Context Learningcited by 886
  7. 2023Large Language Models as Optimizerscited by 830
  8. 2023Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formattingcited by 778
  9. 2023LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attentioncited by 776
  10. 2021P-Tuning v2: Prompt Tuning Can Be Comparable to Fine-tuning Universally Across Scales and Taskscited by 686
  11. 2021Learning To Retrieve Prompts for In-Context Learningcited by 641
  12. 2023Automatic Prompt Optimization with "Gradient Descent" and Beam Searchcited by 628
38 more
  1. 2023Promptbreeder: Self-Referential Self-Improvement Via Prompt Evolutioncited by 521
  2. 2021Recent Advances in Natural Language Processing via Large Pre-Trained Language Models: A Surveycited by 457
  3. 2022Recommendation as Language Processing (RLP): A Unified Pretrain, Personalized Prompt & Predict Paradigm (P5)cited by 396
  4. 2022RLPrompt: Optimizing Discrete Text Prompts with Reinforcement Learningcited by 361
  5. 2021Do Prompt-Based Models Really Understand the Meaning of their Prompts?cited by 324
  6. 2022PromptSource: An Integrated Development Environment and Repository for Natural Language Promptscited by 288
  7. 2022Exploring Visual Prompts for Adapting Large-Scale Modelscited by 271
  8. 2022What does a platypus look like? Generating customized prompts for zero-shot image classificationcited by 263
  9. 2022Black-Box Tuning for Language-Model-as-a-Servicecited by 254
  10. 2023Think before you speak: Training Language Models With Pause Tokenscited by 254
  11. 2023Personalize Segment Anything Model with One Shotcited by 250
  12. 2021SPoT: Better Frozen Model Adaptation through Soft Prompt Transfercited by 235
  13. 2021PPT: Pre-trained Prompt Tuning for Few-shot Learningcited by 232
  14. 2022Demystifying Prompts in Language Models via Perplexity Estimationcited by 215
  15. 2022ST-Adapter: Parameter-Efficient Image-to-Video Transfer Learningcited by 192
  16. 2021CPT: Colorful Prompt Tuning for Pre-trained Vision-Language Modelscited by 182
  17. 2022ZegCLIP: Towards Adapting CLIP for Zero-shot Semantic Segmentationcited by 178
  18. 2024A Survey on Recent Advances in LLM-Based Multi-turn Dialogue Systemscited by 170
  19. 2021Noisy Channel Language Model Prompting for Few-Shot Text Classificationcited by 163
  20. 2021Meta-learning via Language Model In-context Tuningcited by 163
  21. 2022A Review on Language Models as Knowledge Basescited by 156
  22. 2023LM vs LM: Detecting Factual Errors via Cross Examinationcited by 155
  23. 2023Active Prompting with Chain-of-Thought for Large Language Modelscited by 154
  24. 2022Can Machines Help Us Answering Question 16 in Datasheets, and In Turn Reflecting on Inappropriate Content?cited by 150
  25. 2023Large Language Models for Code: Security Hardening and Adversarial Testingcited by 149
  26. 2021Cutting Down on Prompts and Parameters: Simple Few-Shot Learning with Language Modelscited by 139
  27. 2022GrIPS: Gradient-free, Edit-based Instruction Search for Prompting Large Language Modelscited by 127
  28. 2023Automatic Prompt Augmentation and Selection with Chain-of-Thought from Labeled Datacited by 119
  29. 2021Differentiable Prompt Makes Pre-trained Language Models Better Few-shot Learnerscited by 115
  30. 2022CLIP Models are Few-shot Learners: Empirical Studies on VQA and Visual Entailmentcited by 113
  31. 2024Concise Thoughts: Impact of Output Length on LLM Reasoning and Costcited by 108
  32. 2023The Earth is Flat because...: Investigating LLMs' Belief towards Misinformation via Persuasive Conversationcited by 107
  33. 2022Repository-Level Prompt Generation for Large Language Models of Codecited by 106
  34. 2023Text Classification via Large Language Modelscited by 104
  35. 2023Task Contamination: Language Models May Not Be Few-Shot Anymorecited by 98
  36. 2024The Butterfly Effect of Altering Prompts: How Small Changes and Jailbreaks Affect Large Language Model Performancecited by 97
  37. 2023Fuzz4All: Universal Fuzzing with Large Language Modelscited by 88
  38. 2024LISA: Layerwise Importance Sampling for Memory-Efficient Large Language Model Fine-Tuningcited by 88

50 most cited of 376 All 376 (JSON)