Citation Map

Every paper referenced by an arXiv: identifier inside the CircuitKit source tree (src/), mapped to its verified title, authors, arXiv page, and Hugging Face Papers page. Titles were checked against arXiv — if you add a new arXiv: reference in code, add a row here too.

For the broader documentation bibliography (foundational MI papers, tooling), see references.md.

arXiv	Title	Authors	Cited by (module)	Links
2211.00593	Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 Small	Wang, Variengien, Conmy, Shlegeris, Steinhardt	backends/acdc/tasks/ioi_dataset.py	arXiv · HF
2309.16042	Towards Best Practices of Activation Patching in Language Models	Zhang, Nanda	data/corruption/resample.py	arXiv · HF
2403.00745	AtP*: An Efficient and Scalable Method for Localizing LLM Behaviour to Components	Kramár, Lieberum, Shah, Nanda	backends/eap/attribute_node.py	arXiv · HF
2502.04577	Position-aware Automatic Circuit Discovery	Haklay, Orgad, Bau, Mueller, Belinkov	backends/eap/attribute_node.py, data/dataset_schema.py	arXiv · HF
2508.21258	RelP: Faithful and Efficient Circuit Discovery in Language Models via Relevance Patching	Rezaei Jafari, Eberle, Khakzar, Nanda	backends/eap/attribute_node.py	arXiv · HF
2502.06852	EAP-GP: Mitigating Saturation Effect in Gradient-based Automated Circuit Identification	Zhang, Dong, Zhang, Yang, Hu, Liu, Zhou, Wang	backends/eap/attribute_node.py	arXiv · HF
2504.07389	Task-Circuit Quantization: Leveraging Knowledge Localization and Interpretability for Compression	Xiao, Sung, Stengel-Eskin, Bansal	applications/quantization/selectors/tacq_selector.py	arXiv · HF
2604.05876	Mechanistic Circuit-Based Knowledge Editing in Large Language Models	Zhao, He, Zheng, Chen	applications/editing/cake.py, applications/editing/mcircke.py	arXiv · HF
2010.00133	CrowS-Pairs: A Challenge Dataset for Measuring Social Biases in Masked Language Models	Nangia, Vania, Bhalerao, Bowman	data/adapters/pairwise.py	arXiv · HF
2603.23268	SafeSeek: Universal Attribution of Safety Circuits in Language Models	Yu, Fu, Aloqaily, Zhou, Otoum, Fan, Wang, Guo, Wen	data/corruption/benign_rewrite.py	arXiv · HF
2406.11717	Refusal in Language Models Is Mediated by a Single Direction	Arditi, Obeso, Syed, Paleka, Panickssery, Gurnee, Nanda	data/corruption/benign_rewrite.py	arXiv · HF
2312.15710	Alleviating Hallucinations of Large Language Models through Induced Hallucinations	Zhang, Cui, Bi, Shi	data/corruption/llm_counterfactual.py	arXiv · HF
2404.12010	ParaFusion: A Large-Scale LLM-Driven English Paraphrase Dataset	Jayawardena, Yapa	data/corruption/llm_counterfactual.py	arXiv · HF
2305.15054	A Mechanistic Interpretation of Arithmetic Reasoning in Language Models using Causal Mediation Analysis	Stolfo, Belinkov, Sachan	data/corruption/operand_swap.py	arXiv · HF
2004.12265	Causal Mediation Analysis for Interpreting Neural NLP: The Case of Gender Bias	Vig, Gehrmann, Belinkov, Qian, Nevo, Sakenis, Huang, Singer, Shieber	data/corruption/profession_swap.py	arXiv · HF
2411.16105	Adaptive Circuit Behavior and Generalization in Mechanistic Interpretability	Nainani, Vaidyanathan, Yeung, Gupta, Jensen	tasks/builtins/double_io.py, data/task_data/tasks/double_io/	arXiv · HF

References without an arXiv identifier

Title	Authors / Venue	Cited by (module)	Link
IBCircuit: Towards Holistic Circuit Discovery with Information Bottleneck	Bian, Niu, Yuan, Piao, Wu, Huang, Rong, Xu, Cheng, Li — ICML 2025	backends/ibcircuit/, selection/ibcircuit_selector.py	Code

Note: earlier revisions cited arXiv:2408.05520 for IBCircuit; that identifier actually resolves to an unrelated quantum-physics paper ("On stability issues of the HEOM method") and has been corrected to the ICML 2025 reference above. IBCircuit's noise-injection mechanism builds on the information-bottleneck attribution method of Schulz et al. (2020), arXiv:2001.00396.