mlsanityData to 7 Oct 2026

Detecting Conceptual Abstraction in LLMs

arXiv 2404.15848 ↗First posted 24 Apr 2024cs.CL cs.LGcited by 3

Not on the map. Why?

Abstract (arXiv)

We present a novel approach to detecting noun abstraction within a large language model (LLM). Starting from a psychologically motivated set of noun pairs in taxonomic relationships, we instantiate surface patterns indicating hypernymy and analyze the attention matrices produced by BERT. We compare the results to two sets of counterfactuals and show that we can detect hypernymy in the abstraction mechanism, which cannot solely be related to the distributional similarity of noun pairs. Our findings are a first step towards the explainability of conceptual abstraction in LLMs.

Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Builds on 16 of its 35 references that are in the map. First posted on arXiv 24 Apr 2024.

Builds on

16 of 35 references.

References · 16

  1. 2017Attention Is All You Needcited by 59,932
  2. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,204
  3. 2014Convolutional Neural Networks for Sentence Classificationcited by 1,124
  4. 2019What Does BERT Look At? An Analysis of BERT's Attentioncited by 994
  5. 2019Attention is not Explanationcited by 916
  6. 2021Probing Classifiers: Promises, Shortcomings, and Advancescited by 831
  7. 2017Dynamic Routing Between Capsulescited by 689
  8. 2019Attention is not not Explanationcited by 629
  9. 2019Revealing the Dark Secrets of BERTcited by 300
  10. 2018Colorless green recurrent networks dream hierarchicallycited by 210
  11. 2018What Is One Grain of Sand in the Desert? Analyzing Individual Neurons in Deep NLP Modelscited by 124
  12. 2021Neuron-level Interpretation of Deep NLP Models: A Surveycited by 81
4 more
  1. 2019Linguistic generalization and compositionality in modern artificial neural networkscited by 58
  2. 2022Discovering Latent Concepts Learned in BERTcited by 56
  3. 2018NeuroX: A Toolkit for Analyzing Individual Neurons in Neural Networkscited by 31
  4. 2014A Deep Architecture for Semantic Parsingcited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 2; 2026: 1.

3 papers.

  1. 2025Exploring Multimodal Perception in Large Language Models Through Perceptual Strength Ratingscited by 3
  2. 2025Prediction is not Explanation: Revisiting the Explanatory Capacity of Mapping Embeddingscited by 2
  3. 2026Semantic Abstraction for Natural Language Inference: a Methodological Framework for Discovering and Compensating Semantic Knowledge and Reasoning Gaps in Large Language Modelscited by 0