| # | Source (linked) | NeurIPS/ICML/ICLR/Other Reputable Conferences | Reputable journal/news/500+ Citations | Does it satisfy either condition? |
|---|---|---|---|---|
| 1 | Mechanistic Interpretability, Variables, and the Importance of Interpretable Bases | No | No | |
| 2 | Zoom In: An Introduction to Circuits (Distill) | No | Yes, Distill | |
| 3 | A Mathematical Framework for Transformer Circuits | No | 633 Citations | |
| 4 | Mechanistic? · ACL workshop version | Proceedings of the 7th BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks for NLP (Peer-reviewed, archival workshop proceeding at ACL) | No | |
| 5 | The Building Blocks of Interpretability (Distill) | No | Distill | |
| 6 | Transformer Circuits Thread | No | No | |
| 7 | In-context Learning and Induction Heads | No (arXiv) | 600 Citations | |
| 8 | Toy Models of Superposition | No (arXiv / project page) | 368 Citations | |
| 9 | Progress Measures for Grokking via Mechanistic Interpretability | ICLR 2023 (main; notable top 25%) | No | |
| 10 | Sparse Autoencoders Find Highly Interpretable Features in Language Models | ICLR 2024 (main) | No | |
| 11 | Towards Monosemanticity: Decomposing LMs with Dictionary Learning | No | 587 Citations | |
| 12 | This startup wants to reprogram the mind of AI—and just got $50 million to do it | No | Fast Company (news) | |
| 13 | ICML 2024 Mechanistic Interpretability Workshop | ICML 2024 (Workshop; not main proceedings) | No | |
| 14 | Tweet by @ch402 (Chris Olah), Jul 29, 2024 | No | No | |
| 15 | Mechanistic Interpretability Quickstart Guide | No | No | |
| 16 | Mechanistic Interpretability Workshop 2024 (site) | ICML 2024 (Workshop; not main proceedings) | No | |
| 17 | Linguistic Regularities in Continuous Space Word Representations | ACL | 5601 Citations | |
| 18 | The Linear Representation Hypothesis and the Geometry of LLMs | ICML 2024 (main) | No | |
| 19 | Circuits Updates — July 2024: What is a Linear Representation? | No | No | |
| 20 | Open Problems in Mechanistic Interpretability | No (arXiv) | No | |
| 21 | Interpreting Neural Networks through the Polytope Lens | No (arXiv / blog) | No | |
| 22 | RNNs Learn to Store and Generate Sequences using Non-Linear Representations | 7th BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks for NLP (Peer-reviewed, archival workshop proceeding at ACL) | No | |
| 23 | Polysemanticity and Capacity in Neural Networks | No (arXiv) | No | |
| 24 | Mechanistic Interpretability for AI Safety — A Review | No | TMLR (journal) | |
| 25 | Understanding Intermediate Layers using Linear Classifier Probes | ICLR 2017 (Workshop track) | 1126 Citations | |
| 26 | The Geometry of Truth: Emergent Linear Structure in LLM Representations | COLM 2024 | No | |
| 27 | Refusal in Language Models Is Mediated by a Single Direction | NeurIPS 2024 poster | No | |
| 28 | Linear Representations of Sentiment in Large Language Models | No (arXiv) | No | |
| 29 | Open Problems in Mechanistic Interpretability | No (arXiv) — duplicate of #20 | No | |
| 30 | Investigating Gender Bias in Language Models Using Causal Mediation Analysis | NeurIPS 2020 (main) | No | |
| 31 | Locating and Editing Factual Associations in GPT | NeurIPS 2022 (main) | No | |
| 32 | NLI Models Partially Embed Theories of Lexical Entailment and Negation | Proceedings of the Third BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP (Peer-reviewed, archival workshop proceeding at ACL) | No | |
| 33 | Causal Abstraction: A Theoretical Foundation for Mechanistic Interpretability | No | JMLR (journal) | |
| 34 | Interpretability in the Wild: A Circuit for Indirect Object Identification in GPT-2 Small | ICLR 2023 (poster) | No | |
| 35 | Towards Best Practices of Activation Patching in Language Models: Metrics and Methods | ICLR 2024 (main) | No | |
| 36 | Open Problems in Mechanistic Interpretability | No (arXiv) — duplicate of #20 | No | |
| 37 | Attribution Patching: Activation Patching At Industrial Scale | No (blog) | No | |
| 38 | Attribution Patching Outperforms Automated Circuit Discovery | BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks for NLP (Peer-reviewed, archival workshop proceeding at ACL) | No | |
| 39 | AtP*: An efficient and scalable method for localizing LLM behaviour to components | No (arXiv) | No | |
| 40 | Axiomatic Attribution for Deep Networks | ICML 2017 (main) | 8647 Citations | |
| 41 | Open Problems in Mechanistic Interpretability | No (arXiv) — duplicate of #20 | No | |
| 42 | Scaling and evaluating sparse autoencoders | ICLR 2025 (Oral) | No | |
| 43 | Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoders | No (arXiv) | No | |
| 44 | Circuits Updates — January 2025: Dictionary Learning Optimization Techniques | No (blog/tech note) | No |
