mlsanityData to 7 Oct 2026

Towards Mechanistic Interpretability of Graph Transformers via Attention Graphs

arXiv 2502.12352 ↗First posted 17 Feb 2025cs.LG cs.AIcited by 12

Not on the map. Why?

Abstract (arXiv)

We introduce Attention Graphs, a new tool for mechanistic interpretability of Graph Neural Networks (GNNs) and Graph Transformers based on the mathematical equivalence between message passing in GNNs and the self-attention mechanism in Transformers. Attention Graphs aggregate attention matrices across Transformer layers and heads to describe how information flows among input nodes. Through experiments on homophilous and heterophilous node classification tasks, we analyze Attention Graphs from a network science perspective and find that: (1) When Graph Transformers are allowed to learn the optimal graph structure using all-to-all attention among input nodes, the Attention Graphs learned by the model do not tend to correlate with the input/original graph structure; and (2) For heterophilous graphs, different Graph Transformer variants can achieve similar performance while utilising distinct information flow patterns. Open source code: https://github.com/batu-el/understanding-inductive-biases-of-gnns

Key facts. Cited by 12 papers in the map (data through 7 Oct 2026). Builds on 29 of its 46 references that are in the map. First posted on arXiv 17 Feb 2025.

Builds on

29 of 46 references.

References · 29

  1. 2017Attention Is All You Needcited by 59,932
  2. 2016Semi-Supervised Classification with Graph Convolutional Networkscited by 11,282
  3. 2017Graph Attention Networkscited by 8,854
  4. 2018How Powerful are Graph Neural Networks?cited by 4,355
  5. 2017Neural Message Passing for Quantum Chemistrycited by 3,460
  6. 2022Locating and Editing Factual Associations in GPTcited by 2,609
  7. 2016Revisiting Semi-Supervised Learning with Graph Embeddingscited by 970
  8. 2023Vision Transformers Need Registerscited by 844
  9. 2023Progress measures for grokking via mechanistic interpretabilitycited by 822
  10. 2020Geom-GCN: Geometric Graph Convolutional Networkscited by 821
  11. 2019GNNExplainer: Generating Explanations for Graph Neural Networkscited by 735
  12. 2022Recipe for a General, Powerful, Scalable Graph Transformercited by 702
17 more
  1. 2020A Generalization of Transformer Networks to Graphscited by 653
  2. 2019Multi-scale Attributed Node Embeddingcited by 543
  3. 2019MixHop: Higher-Order Graph Convolutional Architectures via Sparsified Neighborhood Mixingcited by 442
  4. 2020Discovering Symbolic Models from Deep Learning with Inductive Biasescited by 335
  5. 2023A critical look at the evaluation of GNNs under heterophily: Are we really making progress?cited by 332
  6. 2020XGNN: Towards Model-Level Explanations of Graph Neural Networkscited by 211
  7. 2024Flex Attention: A Programming Model for Generating Optimized Attention Kernelscited by 180
  8. 2020Compositional Explanations of Neuronscited by 172
  9. 2019Understanding Attention and Generalization in Graph Neural Networkscited by 169
  10. 2023On Over-Squashing in Message Passing Neural Networks: The Impact of Width, Depth, and Topologycited by 163
  11. 2023Artificial Intelligence for Science in Quantum, Atomistic, and Continuum Systemscited by 118
  12. 2023Everything is Connected: Graph Neural Networkscited by 85
  13. 2023Attending to Graph Transformerscited by 81
  14. 2024InterPLM: Discovering Interpretable Features in Protein Language Models via Sparse Autoencoderscited by 64
  15. 2022Characterizing Graph Datasets for Node Classification: Homophily-Heterophily Dichotomy and Beyondcited by 60
  16. 2021GCExplainer: Human-in-the-Loop Concept-based Explanations for Graph Neural Networkscited by 35
  17. 2024Understanding Biology in the Age of Artificial Intelligencecited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 8; 2026: 4.

12 papers.

  1. 2025FAIRGAME: a Framework for AI Agents Bias Recognition using Game Theorycited by 13
  2. 2025Do LLMs trust AI regulation? Emerging behaviour of game-theoretic LLM agentscited by 11
  3. 2025Attention (as Discrete-Time Markov) Chainscited by 11
  4. 2025Neural Message-Passing on Attention Graphs for Hallucination Detectioncited by 5
  5. 2025Strategic Communication and Language Bias in Multi-Agent LLM Coordinationcited by 3
  6. 2026Geometry of Reason: Spectral Signatures of Valid Mathematical Reasoningcited by 3
  7. 2025Task complexity shapes internal representations and robustness in neural networkscited by 2
  8. 2026When Numbers Start Talking: Implicit Numerical Coordination Among LLM-Based Agentscited by 2
  9. 2026Spectral Guardrails for Agents in the Wild: Detecting Tool Use Hallucinations via Attention Topologycited by 2
  10. 2025Chi-Geometry: A Library for Benchmarking Chirality Prediction of GNNscited by 0
  11. 2025DARTS-GT: Differentiable Architecture Search for Graph Transformers with Quantifiable Instance-Specific Interpretability Analysiscited by 0
  12. 2026SLASH the Sink: Sharpening Structural Attention Inside LLMscited by 0