mlsanityData to 14 Oct 2026

What do Toothbrushes do in the Kitchen? How Transformers Think our World is Structured

arXiv 2204.05673 ↗First posted 12 Apr 2022cs.CLcited by 5

On the map: LLM reasoning → Reasoning benchmarks and evaluation

Abstract (arXiv)

Transformer-based models are now predominant in NLP. They outperform approaches based on static models in many respects. This success has in turn prompted research that reveals a number of biases in the language models generated by transformers. In this paper we utilize this research on biases to investigate to what extent transformer-based language models allow for extracting knowledge about object relations (X occurs in Y; X consists of Z; action A involves using X). To this end, we compare contextualized models with their static counterparts. We make this comparison dependent on the application of a number of similarity measures and classifiers. Our results are threefold: Firstly, we show that the models combined with the different similarity measures differ greatly in terms of the amount of knowledge they allow for extracting. Secondly, our results suggest that similarity measures perform much worse than classifier-based approaches. Thirdly, we show that, surprisingly, static models perform almost as well as contextualized models -- in some cases even better.

Key facts. Cited by 5 papers in the map (data through 14 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 40 of its 71 references that are in the map. First posted on arXiv 12 Apr 2022.

Builds on

40 of 71 references.

References · 40

  1. 2017Attention Is All You Needcited by 60,361
  2. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,399
  3. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  4. 2020Language Models are Few-Shot Learnerscited by 32,419
  5. 2014Microsoft COCO: Common Objects in Contextcited by 17,608
  6. 2019RoBERTa: A Robustly Optimized BERT Pretraining Approachcited by 13,372
  7. 2013Distributed Representations of Words and Phrases and their Compositionalitycited by 3,474
  8. 2015ShapeNet: An Information-Rich 3D Model Repositorycited by 2,897
  9. 2019ALBERT: A Lite BERT for Self-supervised Learning of Language Representationscited by 2,790
  10. 2020The Pile: An 800GB Dataset of Diverse Text for Language Modelingcited by 2,411
  11. 2019Language Models as Knowledge Bases?cited by 1,936
  12. 2019What Does BERT Look At? An Analysis of BERT's Attentioncited by 1,003
28 more
  1. 2019How Can We Know What Language Models Know?cited by 959
  2. 2016Semantics derived automatically from language corpora contain human-like biasescited by 901
  3. 2020A Primer in BERTology: What we know about how BERT workscited by 815
  4. 2019How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddingscited by 784
  5. 2019BLiMP: The Benchmark of Linguistic Minimal Pairs for Englishcited by 564
  6. 2018PartNet: A Large-scale Benchmark for Fine-grained and Hierarchical Part-level 3D Object Understandingcited by 526
  7. 2016Semantic Scene Completion from a Single Depth Imagecited by 472
  8. 2019What do you learn from context? Probing for sentence structure in contextualized word representationscited by 438
  9. 2019On Measuring Social Biases in Sentence Encoderscited by 398
  10. 20203D-FRONT: 3D Furnished Rooms with layOuts and semaNTicscited by 380
  11. 2019KG-BERT: BERT for Knowledge Graph Completioncited by 324
  12. 2018Analysis Methods in Neural Language Processing: A Surveycited by 277
  13. 2019Measuring Bias in Contextualized Word Representationscited by 276
  14. 2019Compositionality decomposed: how do neural networks generalise?cited by 265
  15. 2019Visualizing and Measuring the Geometry of BERTcited by 216
  16. 2017Advances in Pre-Training Distributed Word Representationscited by 207
  17. 2020How Context Affects Language Models' Factual Predictionscited by 181
  18. 2020Birds have four legs?! NumerSense: Probing Numerical Commonsense Knowledge of Pre-trained Language Modelscited by 110
  19. 2020Language Models as Knowledge Bases: On Entity Representations, Storage Capacity, and Paraphrased Queriescited by 107
  20. 2019Inducing Relational Knowledge from BERTcited by 75
  21. 2019Do Neural Language Representations Learn Physical Commonsense?cited by 62
  22. 2021Provable Limitations of Acquiring Meaning from Ungrounded Form: What Will Future Language Models Understand?cited by 55
  23. 2020A Comparison of LSTM and BERT for Small Corpuscited by 40
  24. 2015Take and Took, Gaggle and Goose, Book and Read: Evaluating the Utility of Vector Differences for Lexical Relation Learningcited by 29
  25. 2019Cracking the Contextual Commonsense Code: Understanding Commonsense Reasoning Aptitude of Deep Contextual Representationscited by 24
  26. 2021Obtaining Better Static Word Embeddings Using Contextual Embedding Modelscited by 18
  27. 2021Visual Grounding Strategies for Text-Only Natural Language Processingcited by 7
  28. 2019Improving BERT Fine-tuning with Embedding Normalizationcited by 3

Built on by

Cited by year

Citing papers per year: 2022: 2; 2023: 2; 2024: 1.

5 papers.

  1. 2023Weakly Supervised Detection of Hallucinations in LLM Activationscited by 20
  2. 2023Toward Fairness in Text Generation via Mutual Information Minimization based on Importance Samplingcited by 12
  3. 2024Language-Based Depth Hints for Monocular Depth Estimationcited by 3
  4. 2022Where's the Learning in Representation Learning for Compositional Semantics and the Case of Thematic Fitcited by 2
  5. 2022ObjCAViT: Improving Monocular Depth Estimation Using Natural Language Models And Image-Object Cross-Attentioncited by 1

Similar, not cited

Close in wording, no citation either way

  1. 2025Linguistic Interpretability of Transformer-based Language Models: a systematic review
  2. 2026LLMs Explain't: A Post-Mortem on Semantic Interpretability in Transformer Models
  3. 2026Transformer-Based Language Models Across Domain Verticals: Architectures, Applications and Critical Assessment
  4. 2022minicons: Enabling Flexible Behavioral and Representational Analyses of Transformer Language Models
  5. 2025Generalization or Hallucination? Understanding Out-of-Context Reasoning in Transformers
  6. 2021Transformer visualization via dictionary learning: contextualized embedding as a linear superposition of transformer factors
  7. 2023Observable Propagation: Uncovering Feature Vectors in Transformers
  8. 2024Do Efficient Transformers Really Save Computation?