Lune

ICLR2025Top-tier venue

Efficient Automated Circuit Discovery in Transformers using Contextual Decomposition

Aliyah R. Hsu, Georgia Zhou, Yeshwanth Cherapanamjeri, Yaxuan Huang, Anobel Y. Odisho, Peter R. Carroll, Bin Yu

Abstract

Automated mechanistic interpretation research has attracted great interest due to its potential to scale explanations of neural network internals to large models. Existing automated circuit discovery work relies on activation patching or its approximations to identify subgraphs in models for specific tasks (circuits). They often suffer from slow runtime, approximation errors, and specific requirements of metrics, such as non-zero gradients. In this work, we introduce contextual decomposition for transformers (CD-T) to build interpretable circuits in large language models. CD-T can produce circuits at any level of abstraction and is the first to efficiently produce circuits as fine-grained as attention heads at specific sequence positions. CD-T is compatible to all transformer types, and requires no training or manually-crafted examples. CD-T consists of a set of mathematical equations to isolate contribution of model features. Through recursively computing contribution of all nodes in a computational graph of a model using CD-T followed by pruning, we are able to reduce circuit discovery runtime from hours to seconds compared to state-of-the-art baselines. On three standard circuit evaluation datasets (indirect object identification, greater-than comparisons, and docstring completion), we demonstrate that CD-T outperforms ACDC and EAP by better recovering the manual circuits with an average of 97% ROC AUC under low runtimes. In addition, we provide evidence that faithfulness of CD-T circuits is not due to random chance by showing our circuits are 80% more faithful than random circuits of up to 60% of the original model size. Finally, we show CD-T circuits are able to perfectly replicate original models' behavior (faithfulness = 1) using fewer nodes than the baselines for all tasks. Our results underscore the great promise of CD-T for efficient automated mechanistic interpretability, paving the way for new insights into the workings of large language models. All code for using CD-T and reproducing results is made available on Github. 1

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

10 top-tier papers have cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • Towards Automated Circuit Discovery for Mechanistic Interpretability cites Locating and Editing Factual Associations in GPT
  • Towards Automated Circuit Discovery for Mechanistic Interpretability cites Causal Abstractions of Neural Networks
  • The Atlas of In-Context Learning: How Attention Heads Shape In-Context Retrieval Augmentation cites Towards Automated Circuit Discovery for Mechanistic Interpretability
  • The Atlas of In-Context Learning: How Attention Heads Shape In-Context Retrieval Augmentation cites Causal Abstractions of Neural Networks
  • The Atlas of In-Context Learning: How Attention Heads Shape In-Context Retrieval Augmentation cites Language Models are Few-Shot Learners
  • Locating and Editing Factual Associations in GPT cites Investigating Gender Bias in Language Models Using Causal Mediation Analysis
  • Locating and Editing Factual Associations in GPT cites Language Models are Few-Shot Learners
  • Formal Mechanistic Interpretability: Automated Circuit Discovery with Provable Guarantees cites Towards Automated Circuit Discovery for Mechanistic Interpretability
  • Beyond Components: Singular Vector-Based Interpretability of Transformer Circuits cites Towards Automated Circuit Discovery for Mechanistic Interpretability
  • Beyond Components: Singular Vector-Based Interpretability of Transformer Circuits cites Locating and Editing Factual Associations in GPT
  • Beyond Components: Singular Vector-Based Interpretability of Transformer Circuits cites Language Models are Few-Shot Learners
  • Rethinking Circuit Completeness in Language Models: AND, OR, and ADDER Gates cites Towards Automated Circuit Discovery for Mechanistic Interpretability
  • Rethinking Circuit Completeness in Language Models: AND, OR, and ADDER Gates cites Locating and Editing Factual Associations in GPT
  • Rethinking Circuit Completeness in Language Models: AND, OR, and ADDER Gates cites Investigating Gender Bias in Language Models Using Causal Mediation Analysis
  • ProxySPEX: Inference-Efficient Interpretability via Sparse Feature Interactions in LLMs cites Towards Automated Circuit Discovery for Mechanistic Interpretability
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers10

  1. ProxySPEX: Inference-Efficient Interpretability via Sparse Feature Interactions in LLMsLandon Butler, Abhineet Agarwal, Justin Singh Kang, Yigit Efe Erginbas et al.NeurIPS 2025 · 19 citations
  2. The Atlas of In-Context Learning: How Attention Heads Shape In-Context Retrieval AugmentationPatrick Kahardipraja, Reduan Achtibat, Thomas Wiegand, Wojciech Samek et al.NeurIPS 2025 · 13 citations
  3. Rethinking Circuit Completeness in Language Models: AND, OR, and ADDER GatesHang Chen, Jiaying Zhu, Xinyu Yang, Wenya WangNeurIPS 2025 · 11 citations
  4. Formal Mechanistic Interpretability: Automated Circuit Discovery with Provable GuaranteesItamar Hadad, Guy Katz, Shahaf BassanICLR 2026 · 10 citations
  5. Beyond Components: Singular Vector-Based Interpretability of Transformer CircuitsAreeb Ahmad, Abhinav Joshi, Ashutosh ModiNeurIPS 2025 · 9 citations

Builds on13

  1. Language Models are Few-Shot LearnersTom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah et al.NeurIPS 2020 · 64,255 citations
  2. Locating and Editing Factual Associations in GPTKevin Meng, David Bau, Alex Andonian, Yonatan BelinkovNeurIPS 2022 · 3,415 citations
  3. Towards Automated Circuit Discovery for Mechanistic InterpretabilityArthur Conmy, Augustine N. Mavor-Parker, Aengus Lynch, Stefan Heimersheim et al.NeurIPS 2023 · 861 citations
  4. Investigating Gender Bias in Language Models Using Causal Mediation AnalysisJesse Vig, Sebastian Gehrmann, Yonatan Belinkov, Sharon Qian et al.NeurIPS 2020 · 851 citations
  5. Causal Abstractions of Neural NetworksAtticus Geiger, Hanson Lu, Thomas Icard, Christopher PottsNeurIPS 2021 · 516 citations

Related papers

  1. EAP-GP: Mitigating Saturation Effect in Gradient-based Automated Circuit IdentificationLin Zhang, Wenshuo Dong, Zhuoran Zhang, Shu Yang et al.NeurIPS 2025 · 26 citations
  2. Circuit Compositions: Exploring Modular Structures in Transformer-Based Language ModelsPhilipp Mondorf, Sondre Wold, Barbara PlankACL 2025 · 5 citations
  3. Transcoders find interpretable LLM feature circuitsJacob Dunefsky, Philippe Chlenski, Neel NandaNeurIPS 2024 · 222 citations
  4. Circuit Insights: Towards Interpretability Beyond ActivationsElena Golimblevskaia, Aakriti Jain, Bruno Puri, Ammar Ibrahim et al.ICLR 2026 · 4 citations
  5. Towards Global-level Mechanistic Interpretability: A Perspective of Modular Circuits of Large Language ModelsYinhan He, Wendy Zheng, Yushun Dong, Yaochen Zhu et al.ICML 2025
Dusk over the sea between two cliffs drawn in fine vertical lines