mlsanityData to 7 Oct 2026

What BERT is not: Lessons from a new suite of psycholinguistic diagnostics for language models

arXiv 1907.13528 ↗First posted 31 Jul 2019cs.CL cs.AIcited by 317

Not on the map. Why?

Abstract (arXiv)

Pre-training by language modeling has become a popular and successful approach to NLP tasks, but we have yet to understand exactly what linguistic capacities these pre-training processes confer upon models. In this paper we introduce a suite of diagnostics drawn from human language experiments, which allow us to ask targeted questions about the information used by language models for generating predictions in context. As a case study, we apply these diagnostics to the popular BERT model, finding that it can generally distinguish good from bad completions involving shared category or role reversal, albeit with less sensitivity than humans, and it robustly retrieves noun hypernyms, but it struggles with challenging inferences and role-based event prediction -- and in particular, it shows clear insensitivity to the contextual impacts of negation.

Key facts. Cited by 317 papers in the map (data through 7 Oct 2026). First posted on arXiv 31 Jul 2019.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 79; 2022: 74; 2023: 73; 2024: 45; 2025: 23; 2026: 23.

317 papers.

Sort:

By area:

  1. 2021Pre-train, Prompt, and Predict: A Systematic Survey of Prompting Methods in Natural Language Processingcited by 2,222
  2. 2022When and why vision-language models behave like bags-of-words, and what to do about it?cited by 603
  3. 2021Recent Advances in Natural Language Processing via Large Pre-Trained Language Models: A Surveycited by 457
  4. 2021Dynabench: Rethinking Benchmarking in NLPcited by 447
  5. 2022Using cognitive psychology to understand GPT-3cited by 389
  6. 2021Measuring and Improving Consistency in Pretrained Language Modelscited by 361
  7. 2021True Few-Shot Learning with Language Modelscited by 343
  8. 2021AI and the Everything in the Whole Wide World Benchmarkcited by 340
  9. 2021Do Prompt-Based Models Really Understand the Meaning of their Prompts?cited by 324
  10. 2021Pre-Trained Models: Past, Present and Futurecited by 319
  11. 2023Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Counterfactual Taskscited by 243
  12. 2021Masked Language Modeling and the Distributional Hypothesis: Order Word Matters Pre-training for Littlecited by 176
38 more
  1. 2022A Review on Language Models as Knowledge Basescited by 156
  2. 2024Have Faith in Faithfulness: Going Beyond Circuit Overlap When Finding Model Mechanismscited by 144
  3. 2023AI Transparency in the Age of LLMs: A Human-Centered Research Roadmapcited by 119
  4. 2023Are Emergent Abilities in Large Language Models just In-Context Learning?cited by 109
  5. 2021Knowledgeable or Educated Guess? Revisiting Language Models as Knowledge Basescited by 107
  6. 2023The Generative AI Paradox: "What It Can Create, It May Not Understand"cited by 97
  7. 2023Language Model Behavior: A Comprehensive Surveycited by 92
  8. 2021QA Dataset Explosion: A Taxonomy of NLP Resources for Question Answering and Reading Comprehensioncited by 91
  9. 2022The Fallacy of AI Functionalitycited by 88
  10. 2022MuLan: A Joint Embedding of Music Audio and Natural Languagecited by 85
  11. 2022A very preliminary analysis of DALL-E 2cited by 82
  12. 2023Negative Object Presence Evaluation (NOPE) to Measure Object Hallucination in Vision-Language Modelscited by 80
  13. 2023Language models are not naysayers: An analysis of language models on negation benchmarkscited by 73
  14. 2021Chess as a Testbed for Language Model State Trackingcited by 69
  15. 2021Superbizarre Is Not Superb: Derivational Morphology Improves BERT's Interpretation of Complex Wordscited by 63
  16. 2022minicons: Enabling Flexible Behavioral and Representational Analyses of Transformer Language Modelscited by 58
  17. 2022Probing for the Usage of Grammatical Numbercited by 57
  18. 2022Measuring Causal Effects of Data Statistics on Language Model's `Factual' Predictionscited by 57
  19. 2021BERT is to NLP what AlexNet is to CV: Can Pre-Trained Language Models Identify Analogies?cited by 56
  20. 2021BeliefBank: Adding Memory to a Pre-Trained Language Model for a Systematic Notion of Beliefcited by 56
  21. 2022Can Large Language Models Truly Understand Prompts? A Case Study with Negated Promptscited by 56
  22. 2021Changing the World by Changing the Datacited by 51
  23. 2023Structured prompt interrogation and recursive extraction of semantics (SPIRES): A method for populating knowledge bases using zero-shot learningcited by 50
  24. 2021PROST: Physical Reasoning of Objects through Space and Timecited by 48
  25. 2022On the Blind Spots of Model-Based Evaluation Metrics for Text Generationcited by 48
  26. 2021Differentiable Subset Pruning of Transformer Headscited by 46
  27. 2021Mengzi: Towards Lightweight yet Ingenious Pre-trained Models for Chinesecited by 45
  28. 2023Testing AI on language comprehension tasks reveals insensitivity to underlying meaningcited by 45
  29. 2021Structural Persistence in Language Models: Priming as a Window into Abstract Language Representationscited by 44
  30. 2022Event knowledge in large language models: the gap between the impossible and the unlikelycited by 44
  31. 2024Fantastic Semantics and Where to Find Them: Investigating Which Layers of Generative LLMs Reflect Lexical Semanticscited by 43
  32. 2021Word Acquisition in Neural Language Modelscited by 41
  33. 2022Transformer Grammars: Augmenting Transformer Language Models with Syntactic Inductive Biases at Scalecited by 40
  34. 2022SERENGETI: Massively Multilingual Language Models for Africacited by 40
  35. 2022An Analysis of Negation in Natural Language Understanding Corporacited by 39
  36. 2022Commonsense Knowledge Reasoning and Generation with Pre-trained Language Models: A Surveycited by 38
  37. 2022CONDAQA: A Contrastive Reading Comprehension Dataset for Reasoning about Negationcited by 38
  38. 2023Visual cognition in multimodal large language modelscited by 38

50 most cited of 317 All 317 (JSON)