mlsanityData to 14 Oct 2026

Discovering the Hidden Vocabulary of DALLE-2

arXiv 2206.00169 ↗First posted 1 Jun 2022cs.LG cs.CL cs.CRcited by 43

Not on the map. Why?

Abstract (arXiv)

We discover that DALLE-2 seems to have a hidden vocabulary that can be used to generate images with absurd prompts. For example, it seems that textttApoploe vesrreaitais means birds and textttContarra ccetnxniams luryca tanniounons (sometimes) means bugs or pests. We find that these prompts are often consistent in isolation but also sometimes in combinations. We present our black-box method to discover words that seem random but have some correspondence to visual concepts. This creates important security and interpretability challenges.

Key facts. Cited by 43 papers in the map (data through 14 Oct 2026). Builds on 5 of its 6 references that are in the map. First posted on arXiv 1 Jun 2022.

Builds on

5 of 6 references.

References · 5

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2022Hierarchical Text-Conditional Image Generation with CLIP Latentscited by 6,585
  3. 2022Photorealistic Text-to-Image Diffusion Models with Deep Language Understandingcited by 6,355
  4. 2021Zero-Shot Text-to-Image Generationcited by 4,288
  5. 2022A very preliminary analysis of DALL-E 2cited by 82

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2022: 4; 2023: 21; 2024: 8; 2025: 8; 2026: 2.

43 papers.

Sort:

By area:

  1. 2023Hard Prompts Made Easy: Gradient-Based Discrete Optimization for Prompt Tuning and Discoverycited by 336
  2. 2023TextDiffuser-2: Unleashing the Power of Language Models for Text Renderingcited by 140
  3. 2023LEGO-Prover: Neural Theorem Proving with Growing Librariescited by 109
  4. 2023ChatGPT is not all you need. A State of the Art Review of large Generative AI modelscited by 86
  5. 2023A Pilot Study of Query-Free Adversarial Attack against Stable Diffusioncited by 76
  6. 2024Jailbreaking Prompt Attack: A Controllable Adversarial Attack against Diffusion Modelscited by 72
  7. 2024Latent Guard: a Safety Framework for Text-to-image Generationcited by 71
  8. 2025Safety at Scale: A Comprehensive Survey of Large Model and Agent Safetycited by 67
  9. 2023Black Box Adversarial Prompting for Foundation Modelscited by 62
  10. 2023Large AI Models in Health Informatics: Applications, Challenges, and the Futurecited by 61
  11. 2023SurrogatePrompt: Bypassing the Safety Filter of Text-to-Image Models via Substitutioncited by 34
  12. 2022Adversarial Attacks on Image Generation With Made-Up Wordscited by 32
31 more
  1. 2022DALL-E 2 Fails to Reliably Capture Common Syntactic Processescited by 31
  2. 2023Discovering Failure Modes of Text-guided Diffusion Models via Adversarial Searchcited by 25
  3. 2023Generative Watermarking Against Unauthorized Subject-Driven Image Synthesiscited by 23
  4. 2023Generative Models as a Complex Systems Science: How can we make sense of large language model behavior?cited by 19
  5. 2023Online Advertisements with LLMs: Opportunities and Challengescited by 17
  6. 2023Benchmarking Zero-Shot Recognition with Vision-Language Models: Challenges on Granularity and Specificitycited by 14
  7. 2022DALLE-URBAN: Capturing the urban design expertise of large text to image transformerscited by 12
  8. 2024UPAM: Unified Prompt Attack in Text-to-Image Generation Models Against Both Textual Filters and Visual Checkerscited by 12
  9. 2023StraIT: Non-autoregressive Generation with Stratified Image Transformercited by 10
  10. 2023ChatGPT in the Age of Generative AI and Large Language Models: A Concise Surveycited by 10
  11. 2025Rethinking Layered Graphic Design Generation with a Top-Down Approachcited by 10
  12. 2024A Survey on Responsible Generative AI: What to Generate and What Notcited by 8
  13. 2023Prompts have evil twinscited by 7
  14. 2024Global-Local Image Perceptual Score (GLIPS): Evaluating Photorealistic Quality of AI-Generated Imagescited by 7
  15. 2025Unified Prompt Attack Against Text-to-Image Generation Modelscited by 5
  16. 2025Why is prompting hard? Understanding prompts on binary sequence predictorscited by 4
  17. 2023Gradient-Based Word Substitution for Obstinate Adversarial Examples Generation in Language Modelscited by 3
  18. 2023On quantifying and improving realism of images generated with diffusioncited by 3
  19. 2023Hypernymy Understanding Evaluation of Text-to-Image Models via WordNet Hierarchycited by 3
  20. 2023The Challenges of Image Generation Models in Generating Multi-Component Imagescited by 3
  21. 2024Discovering Universal Semantic Triggers for Text-to-Image Synthesiscited by 3
  22. 2024DREAM: Improving Video-Text Retrieval Through Relevance-Based Augmentation Using Large Foundation Modelscited by 3
  23. 2024Evaluating Durability: Benchmark Insights into Multimodal Watermarkingcited by 3
  24. 2023Fooling Contrastive Language-Image Pre-trained Models with CLIPMasterPrintscited by 2
  25. 2025QPM: Discrete Optimization for Globally Interpretable Image Classificationcited by 2
  26. 2025SceneVTG++: Controllable Multilingual Visual Text Generation in the Wildcited by 1
  27. 2025An Exploration of Default Images in Text-to-Image Generationcited by 1
  28. 2022Implementing and Experimenting with Diffusion Models for Text-to-Image Generationcited by 0
  29. 2025FoCLIP: A Feature-Space Misalignment Framework for CLIP-Based Image Manipulation and Detectioncited by 0
  30. 2026Position: Preventing AI-Generated CSAM Necessitates New Approaches to AI Safetycited by 0
  31. 2026Mind the Gap: Zero-Query Jailbreaks via Filter-Generator Discrepancy in Text-to-Image Systemscited by 0