mlsanityData to 14 Oct 2026

Circuit Component Reuse Across Tasks in Transformer Language Models

arXiv 2310.08744 ↗First posted 12 Oct 2023cs.CL cs.LGcited by 108

On the map: Mechanistic interpretability → Circuits and mechanistic analysis

Abstract (arXiv)

Recent work in mechanistic interpretability has shown that behaviors in language models can be successfully reverse-engineered through circuit analysis. A common criticism, however, is that each circuit is task-specific, and thus such analysis cannot contribute to understanding the models at a higher level. In this work, we present evidence that insights (both low-level findings about specific heads and higher-level findings about general algorithms) can indeed generalize across tasks. Specifically, we study the circuit discovered in Wang et al. (2022) for the Indirect Object Identification (IOI) task and 1.) show that it reproduces on a larger GPT2 model, and 2.) that it is mostly reused to solve a seemingly different task: Colored Objects (Ippolito & Callison-Burch, 2023). We provide evidence that the process underlying both tasks is functionally very similar, and contains about a 78% overlap in in-circuit attention heads. We further present a proof-of-concept intervention experiment, in which we adjust four attention heads in middle layers in order to 'repair' the Colored Objects circuit and make it behave like the IOI circuit. In doing so, we boost accuracy from 49.6% to 93.7% on the Colored Objects task and explain most sources of error. The intervention affects downstream attention heads in specific ways predicted by their interactions in the IOI circuit, indicating that this subcircuit behavior is invariant to the different task inputs. Overall, our results provide evidence that it may yet be possible to explain large language models' behavior in terms of a relatively small number of interpretable task-general algorithmic building blocks and computational components.

Key facts. Cited by 108 papers in the map (data through 14 Oct 2026). Placed in Circuits and mechanistic analysis, one of 13 directions in Mechanistic interpretability. Builds on 20 of its 29 references that are in the map. First posted on arXiv 12 Oct 2023.

Builds on

20 of 29 references.

Closest ancestors · 3

Same direction, newest first

  1. 2023Towards Automated Circuit Discovery for Mechanistic Interpretability28 Apr
  2. 2023Progress measures for grokking via mechanistic interpretability12 Jan
  3. 2022Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small1 Nov

Other references · 17

  1. 2022Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language modelscited by 1,998
  2. 2020Transformer Feed-Forward Layers Are Key-Value Memoriescited by 1,336
  3. 2022Mass-Editing Memory in a Transformercited by 1,001
  4. 2022In-context Learning and Induction Headscited by 964
  5. 2019Analyzing Multi-Head Self-Attention: Specialized Heads Do the Heavy Lifting, the Rest Can Be Prunedcited by 856
  6. 2023Dissecting Recall of Factual Associations in Auto-Regressive Language Modelscited by 528
  7. 2023Finding Neurons in a Haystack: Case Studies with Sparse Probingcited by 334
  8. 2019A Multiscale Visualization of Attention in the Transformer Modelcited by 292
  9. 2013Direct and Indirect Effectscited by 280
  10. 2023How does GPT-2 compute greater-than?: Interpreting mathematical abilities in a pre-trained language modelcited by 202
  11. 2023Localizing Model Behavior with Path Patchingcited by 195
  12. 2023Does Circuit Analysis Interpretability Scale? Evidence from Multiple Choice Capabilities in Chinchillacited by 149
5 more
  1. 2018Learning long-range spatial dependencies with horizontal gated-recurrent unitscited by 98
  2. 2023Language Models Implement Simple Word2Vec-style Vector Arithmeticcited by 98
  3. 2020Investigating Transferability in Pretrained Language Modelscited by 28
  4. 2023Generative Models as a Complex Systems Science: How can we make sense of large language model behavior?cited by 19
  5. 2022What Changed? Investigating Debiasing Methods using Causal Mediation Analysiscited by 7

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 2; 2024: 45; 2025: 39; 2026: 22.

108 papers.

Sort:

By area:

  1. 2024Assessing the Brittleness of Safety Alignment via Pruning and Low-Rank Modificationscited by 231
  2. 2024How to use and interpret activation patchingcited by 214
  3. 2024Have Faith in Faithfulness: Going Beyond Circuit Overlap When Finding Model Mechanismscited by 149
  4. 2024The Remarkable Robustness of LLMs: Stages of Inference?cited by 146
  5. 2024A Practical Review of Mechanistic Interpretability for Transformer-Based Language Modelscited by 118
  6. 2024Round and Round We Go! What makes Rotary Positional Encodings useful?cited by 115
  7. 2024Rethinking Interpretability in the Era of Large Language Modelscited by 111
  8. 2024AtP*: An efficient and scalable method for localizing LLM behaviour to componentscited by 111
  9. 2024Universal Neurons in GPT2 Language Modelscited by 102
  10. 2024A Primer on the Inner Workings of Transformer-based Language Modelscited by 85
  11. 2024Attention Heads of Large Language Models: A Surveycited by 67
  12. 2024LLM Circuit Analyses Are Consistent Across Training and Scalecited by 63
38 more
  1. 2024Knowledge Circuits in Pretrained Transformerscited by 60
  2. 2025MIB: A Mechanistic Interpretability Benchmarkcited by 56
  3. 2024A Mechanistic Analysis of a Transformer Trained on a Symbolic Multi-Step Reasoning Taskcited by 49
  4. 2024Talking Heads: Understanding Inter-layer Communication in Transformer Language Modelscited by 41
  5. 2025Shared Global and Local Geometry of Language Model Embeddingscited by 36
  6. 2024Answer, Assemble, Ace: Understanding How LMs Answer Multiple Choice Questionscited by 35
  7. 2024Activation Scaling for Steering and Interpreting Language Modelscited by 35
  8. 2025Same Task, Different Circuits: Disentangling Modality-Specific Mechanisms in VLMscited by 35
  9. 2025(How) Do Language Models Track State?cited by 34
  10. 2024Interpreting Key Mechanisms of Factual Recall in Transformer-Based Language Modelscited by 31
  11. 2024What Do VLMs NOTICE? A Mechanistic Interpretability Pipeline for Gaussian-Noise-free Text-Image Corruption and Evaluationcited by 30
  12. 2024Out-of-distribution generalization via composition: a lens through induction heads in Transformerscited by 28
  13. 2024The Same But Different: Structural Similarities and Differences in Multilingual Language Modelingcited by 28
  14. 2025Mechanistic Interpretability of Emotion Inference in Large Language Modelscited by 27
  15. 2024Optimal ablation for interpretabilitycited by 25
  16. 2024Mechanistic?cited by 25
  17. 2024The Quest for the Right Mediator: Surveying Mechanistic Interpretability Through the Lens of Causal Mediation Analysiscited by 24
  18. 2024Learned feature representations are biased by complexity, learning order, position, and morecited by 23
  19. 2024Repetition Neurons: How Do Language Models Produce Repetitions?cited by 23
  20. 2025How Do LLMs Acquire New Knowledge? A Knowledge Circuits Perspective on Continual Pre-Trainingcited by 23
  21. 2025Re-evaluating Theory of Mind evaluation in large language modelscited by 23
  22. 2023An Analysis and Mitigation of the Reversal Cursecited by 22
  23. 2025EAP-GP: Mitigating Saturation Effect in Gradient-based Automated Circuit Identificationcited by 21
  24. 2025Paths Not Taken: Understanding and Mending the Multilingual Factual Recall Pipelinecited by 21
  25. 2024Efficiently Democratizing Medical LLMs for 50 Languages via a Mixture of Language Family Expertscited by 17
  26. 2023Towards Interpretable Sequence Continuation: Analyzing Shared Circuits in Large Language Modelscited by 15
  27. 2024Anchored Answers: Unravelling Positional Bias in GPT-2's Multiple-Choice Questionscited by 15
  28. 2024Understanding the Language Model to Solve the Symbolic Multi-Step Reasoning Problem from the Perspective of Buffer Mechanismcited by 15
  29. 2025Does Time Have Its Place? Temporal Heads: Where Language Models Recall Time-specific Informationcited by 15
  30. 2025Understanding the Skill Gap in Recurrent Language Models: The Role of the Gather-and-Aggregate Mechanismcited by 15
  31. 2024Beyond the Doors of Perception: Vision Transformers Represent Relations Between Objectscited by 13
  32. 2025LLMs as a synthesis between symbolic and distributed approaches to languagecited by 13
  33. 2024Position: An Inner Interpretability Framework for AI Inspired by Lessons from Cognitive Neurosciencecited by 11
  34. 2024Circuit Compositions: Exploring Modular Structures in Transformer-Based Language Modelscited by 11
  35. 2024Understanding Multimodal LLMs: the Mechanistic Interpretability of Llava in Visual Question Answeringcited by 11
  36. 2025Are formal and functional linguistic mechanisms dissociated in language models?cited by 11
  37. 2025Beyond Induction Heads: In-Context Meta Learning Induces Multi-Phase Circuit Emergencecited by 10
  38. 2026LLMs Know They're Wrong and Agree Anyway: The Shared Sycophancy-Lying Circuitcited by 10

50 most cited of 108 All 108 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025Emergence of Minimal Circuits for Indirect Object Identification in Attention-Only Transformers
  2. 2024Adaptive Circuit Behavior and Generalization in Mechanistic Interpretability
  3. 2026Spectral Probe-Circuits: A Three-Step Recipe for Identifying Attention-Head Circuits in Pretrained Transformers
  4. 2026Finding Interpretable Prompt-Specific Circuits in Language Models
  5. 2025Causal Head Gating: A Framework for Interpreting Roles of Attention Heads in Transformers
  6. 2026Quantifying LLM Attention-Head Stability: Implications for Circuit Universality
  7. 2026From Correlation to Cause: A Five-Stage Methodology for Feature Analysis in Transformer Language Models
  8. 2025Discovering Transformer Circuits via a Hybrid Attribution and Pruning Framework