mlsanityData to 7 Oct 2026

Attribution Patching Outperforms Automated Circuit Discovery

arXiv 2310.10348 ↗First posted 16 Oct 2023cs.LG cs.AI cs.CLcited by 196

Not on the map. Why?

Abstract (arXiv)

Automated interpretability research has recently attracted attention as a potential research direction that could scale explanations of neural network behavior to large models. Existing automated circuit discovery work applies activation patching to identify subnetworks responsible for solving specific tasks (circuits). In this work, we show that a simple method based on attribution patching outperforms all existing methods while requiring just two forward passes and a backward pass. We apply a linear approximation to activation patching to estimate the importance of each edge in the computational subgraph. Using this approximation, we prune the least important edges of the network. We survey the performance and limitations of this method, finding that averaged over all tasks our method has greater AUC from circuit recovery than other methods.

Key facts. Cited by 196 papers in the map (data through 7 Oct 2026). Builds on 15 of its 21 references that are in the map. First posted on arXiv 16 Oct 2023.

Builds on

15 of 21 references.

References · 15

  1. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,160
  2. 2022Locating and Editing Factual Associations in GPTcited by 2,609
  3. 2022Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 smallcited by 1,130
  4. 2023Progress measures for grokking via mechanistic interpretabilitycited by 822
  5. 2019Are Sixteen Heads Really Better than One?cited by 803
  6. 2023Towards Automated Circuit Discovery for Mechanistic Interpretabilitycited by 700
  7. 2020What is the State of Neural Network Pruning?cited by 578
  8. 2017Learning Sparse Neural Networks through L0 Regularizationcited by 571
  9. 2021Causal Abstractions of Neural Networkscited by 401
  10. 2023A Survey on Model Compression for Large Language Modelscited by 263
  11. 2023How does GPT-2 compute greater-than?: Interpreting mathematical abilities in a pre-trained language modelcited by 199
  12. 2020Causal Mediation Analysis for Interpreting Neural NLP: The Case of Gender Biascited by 183
3 more
  1. 2023Does Circuit Analysis Interpretability Scale? Evidence from Multiple Choice Capabilities in Chinchillacited by 145
  2. 2021Low-Complexity Probing via Finding Subnetworkscited by 63
  3. 2020An overview of 11 proposals for building safe advanced AIcited by 18

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 31; 2025: 62; 2026: 103.

196 papers.

Sort:

By area:

  1. 2024Mechanistic Interpretability for AI Safety -- A Reviewcited by 415
  2. 2024Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Modelscited by 412
  3. 2024Not All Language Model Features Are One-Dimensionally Linearcited by 201
  4. 2024A Practical Review of Mechanistic Interpretability for Transformer-Based Language Modelscited by 113
  5. 2024Arithmetic Without Algorithms: Language Models Solve Math With a Bag of Heuristicscited by 94
  6. 2024Information Flow Routes: Automatically Interpreting Language Models at Scalecited by 92
  7. 2024A Primer on the Inner Workings of Transformer-based Language Modelscited by 83
  8. 2025Understanding Reasoning in Thinking Language Models via Steering Vectorscited by 69
  9. 2024Knowledge Circuits in Pretrained Transformerscited by 60
  10. 2024LLM Circuit Analyses Are Consistent Across Training and Scalecited by 60
  11. 2024Finding Transformer Circuits with Edge Pruningcited by 58
  12. 2025Large Language Models Share Representations of Latent Grammatical Concepts Across Typologically Diverse Languagescited by 57
38 more
  1. 2025MIB: A Mechanistic Interpretability Benchmarkcited by 49
  2. 2024Answer, Assemble, Ace: Understanding How LMs Answer Multiple Choice Questionscited by 35
  3. 2024Activation Scaling for Steering and Interpreting Language Modelscited by 33
  4. 2024Hypothesis Testing the Circuit Hypothesis in LLMscited by 29
  5. 2025Training Language Models to Explain Their Own Computationscited by 27
  6. 2025Understanding Refusal in Language Models with Sparse Autoencoderscited by 26
  7. 2024InterpBench: Semi-Synthetic Transformers for Evaluating Mechanistic Interpretability Techniquescited by 24
  8. 2024On Behalf of the Stakeholders: Trends in NLP Model Interpretability in the Era of LLMscited by 24
  9. 2024The Quest for the Right Mediator: Surveying Mechanistic Interpretability Through the Lens of Causal Mediation Analysiscited by 24
  10. 2025How Do LLMs Acquire New Knowledge? A Knowledge Circuits Perspective on Continual Pre-Trainingcited by 22
  11. 2026Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Modelscited by 22
  12. 2024Missed Causes and Ambiguous Effects: Counterfactuals Pose Challenges for Interpreting Neural Networkscited by 21
  13. 2025EAP-GP: Mitigating Saturation Effect in Gradient-based Automated Circuit Identificationcited by 21
  14. 2025Towards Understanding Fine-Tuning Mechanisms of LLMs via Circuit Analysiscited by 20
  15. 2024Opening the AI black box: program synthesis via mechanistic interpretabilitycited by 19
  16. 2025Sparse Autoencoder Features for Classifications and Transferabilitycited by 19
  17. 2025Representation Engineering for Large-Language Models: Survey and Research Challengescited by 18
  18. 2026Language Model Circuits Are Sparse in the Neuron Basiscited by 18
  19. 2024A Philosophical Introduction to Language Models - Part II: The Way Forwardcited by 17
  20. 2025ProxySPEX: Inference-Efficient Interpretability via Sparse Feature Interactions in LLMscited by 16
  21. 2025Position-aware Automatic Circuit Discoverycited by 14
  22. 2025Dissecting Bias in LLMs: A Mechanistic Interpretability Perspectivecited by 14
  23. 2025Stochastic Parameter Decompositioncited by 14
  24. 2025Mechanistic Interpretability as Statistical Estimation: A Variance Analysiscited by 14
  25. 2024Sparse Autoencoders Enable Scalable and Reliable Circuit Identification in Language Modelscited by 13
  26. 2024Gumbel Counterfactual Generation From Language Modelscited by 13
  27. 2025When Reasoning Meets Compression: Understanding the Effects of LLMs Compression on Large Reasoning Modelscited by 12
  28. 2025Beyond Transcription: Mechanistic Interpretability in ASRcited by 12
  29. 2024Unveiling Concept Attribution in Diffusion Modelscited by 11
  30. 2025Mechanistic Unveiling of Transformer Circuits: Self-Influence as a Key to Model Reasoningcited by 11
  31. 2025Are formal and functional linguistic mechanisms dissociated in language models?cited by 11
  32. 2025The Atlas of In-Context Learning: How Attention Heads Shape In-Context Retrieval Augmentationcited by 11
  33. 2025The Validation Gap: A Mechanistic Analysis of How Language Models Compute Arithmetic but Fail to Validate Itcited by 10
  34. 2025Contextualize-then-Aggregate: Circuits for In-Context Learning in Gemma-2 2Bcited by 10
  35. 2025Interpretation Meets Safety: A Survey on Interpretation Methods and Tools for Improving LLM Safetycited by 10
  36. 2025RelP: Faithful and Efficient Circuit Discovery in Language Models via Relevance Patchingcited by 10
  37. 2025The Dead Salmons of AI Interpretabilitycited by 10
  38. 2026Formal Mechanistic Interpretability: Automated Circuit Discovery with Provable Guaranteescited by 10

50 most cited of 196 All 196 (JSON)