mlsanityData to 7 Oct 2026

Automatically Identifying Local and Global Circuits with Linear Computation Graphs

arXiv 2405.13868 ↗First posted 22 May 2024cs.LG cs.CLcited by 22

Not on the map. Why?

Abstract (arXiv)

Circuit analysis of any certain model behavior is a central task in mechanistic interpretability. We introduce our circuit discovery pipeline with Sparse Autoencoders (SAEs) and a variant called Transcoders. With these two modules inserted into the model, the model's computation graph with respect to OV and MLP circuits becomes strictly linear. Our methods do not require linear approximation to compute the causal effect of each node. This fine-grained graph identifies both end-to-end and local circuits accounting for either logits or intermediate features. We can scalably apply this pipeline with a technique called Hierarchical Attribution. We analyze three kinds of circuits in GPT-2 Small: bracket, induction, and Indirect Object Identification circuits. Our results reveal new findings underlying existing discoveries.

Key facts. Cited by 22 papers in the map (data through 7 Oct 2026). Builds on 24 of its 40 references that are in the map. First posted on arXiv 22 May 2024.

Builds on

24 of 40 references.

References · 24

  1. 2017Attention Is All You Needcited by 59,932
  2. 2022Locating and Editing Factual Associations in GPTcited by 2,609
  3. 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 1,359
  4. 2017Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)cited by 1,309
  5. 2016Understanding intermediate layers using linear classifier probescited by 1,279
  6. 2023Inference-Time Intervention: Eliciting Truthful Answers from a Language Modelcited by 1,259
  7. 2023Representation Engineering: A Top-Down Approach to AI Transparencycited by 1,239
  8. 2022Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 smallcited by 1,130
  9. 2022In-context Learning and Induction Headscited by 930
  10. 2022Toy Models of Superpositioncited by 928
  11. 2023Towards Automated Circuit Discovery for Mechanistic Interpretabilitycited by 700
  12. 2024Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Modelscited by 412
12 more
  1. 2023Finding Neurons in a Haystack: Case Studies with Sparse Probingcited by 328
  2. 2023In-Context Learning Creates Task Vectorscited by 302
  3. 2023Causal Abstraction: A Theoretical Foundation for Mechanistic Interpretabilitycited by 194
  4. 2024Improving Dictionary Learning with Gated Sparse Autoencoderscited by 171
  5. 2023The Hydra Effect: Emergent Self-repair in Language Model Computationscited by 145
  6. 2021Transformer visualization via dictionary learning: contextualized embedding as a linear superposition of transformer factorscited by 106
  7. 2024AtP*: An efficient and scalable method for localizing LLM behaviour to componentscited by 100
  8. 2017SPINE: SParse Interpretable Neural Embeddingscited by 55
  9. 2024Identifying Semantic Induction Heads to Understand In-Context Learningcited by 46
  10. 2024Locating and Editing Factual Associations in Mambacited by 35
  11. 2024Dictionary Learning Improves Patch-Free Circuit Discovery in Mechanistic Interpretability: A Case Study on Othello-GPTcited by 27
  12. 2019Word Embedding Visualization Via Dictionary Learningcited by 11

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 4; 2025: 11; 2026: 7.

22 papers.

Sort:

By area:

  1. 2024Transcoders Find Interpretable LLM Feature Circuitscited by 186
  2. 2024Llama Scope: Extracting Millions of Features from Llama-3.1-8B with Sparse Autoencoderscited by 133
  3. 2026When Models Manipulate Manifolds: The Geometry of a Counting Taskcited by 43
  4. 2025Harnessing the Reasoning Economy: A Survey of Efficient Reasoning for Large Language Modelscited by 35
  5. 2024Towards Universality: Studying Mechanistic Similarity Across Language Model Architecturescited by 31
  6. 2024Bilinear MLPs enable weight-based mechanistic interpretabilitycited by 28
  7. 2026Language Model Circuits Are Sparse in the Neuron Basiscited by 18
  8. 2025Position-aware Automatic Circuit Discoverycited by 14
  9. 2025Analyze Feature Flow to Enhance Interpretation and Steering in Language Modelscited by 12
  10. 2025Towards Combinatorial Interpretability of Neural Computationcited by 9
  11. 2025Towards Understanding the Nature of Attention with Low-Rank Sparse Decompositioncited by 9
  12. 2025Evolution of Concepts in Language Model Pre-Trainingcited by 8
10 more
  1. 2025Comprehension Without Competence: Architectural Limits of LLMs in Symbolic Computation and Reasoningcited by 7
  2. 2025Dimensional Collapse in Transformer Attention Outputs: A Challenge for Sparse Dictionary Learningcited by 6
  3. 2025Unboxing the Black Box: Mechanistic Interpretability for Algorithmic Understanding of Neural Networkscited by 6
  4. 2025Circuit Insights: Towards Interpretability Beyond Activationscited by 4
  5. 2026Decomposing Query-Key Feature Interactions Using Contrastive Covariancescited by 3
  6. 2026Tracing the Thought of a Grandmaster-level Chess-Playing Transformercited by 2
  7. 2025K-MSHC: Unmasking Minimally Sufficient Head Circuits in Large Language Models with Experiments on Syntactic Classification Taskscited by 1
  8. 2026Hierarchical Sparse Circuit Extraction from Billion-Parameter Language Models through Scalable Attribution Graph Decompositioncited by 0
  9. 2026Automated Attention Pattern Discovery at Scale in Large Language Modelscited by 0
  10. 2026Not How Many, But Which: Parameter Placement in Low-Rank Adaptationcited by 0