🧑‍🔬 Arthur Conmy

🏛️ Affiliations

🎓Stanford UniversityStanford · United States2026 · 1 papercurrent🎓University of TorontoToronto · Canada2026 · 1 papercurrent🎓University of CambridgeCambridge · United Kingdom2021–2023 · 3 papers

From the bylines of their papers (OpenAlex, 2026-10-14), ORCID and Wikidata; institutions on papers are matched by machine. 1 more institution on single bylines are not listed.

🧭 Fields, topics and concepts

Of its works in our library, with the number of works on each.

📅 Works by year

2020s

📄 Papers

The most cited first (OpenAlex and our library).

1. Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2 (2024 · 115 citations)
   Sparse autoencoders (SAEs) break a language model's internal activations into many sparse, often human-interpretable directions.
2. Attribution Patching Outperforms Automated Circuit Discovery (2023 · 81 citations)
   Mechanistic interpretability tries to find circuits, which are small subgraphs of a network that account for a specific behavior. Automated tools such as ACDC rely on activation patching. That method needs an extra forward pass for each edge it tests, so it scales poorly to large models.
3. Improving Dictionary Learning with Gated Sparse Autoencoders (2024 · 42 citations)
   Sparse autoencoders (SAEs) are used to break language model activations into sparse combinations of interpretable directions.

All works

9 works of theirs in our library, the most cited first; 3 with a page here, the others link their original until we summarise them.

1. Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small ↗arXiv:2211.00593 · no summary yetKevin Wang et al.2022 · 376 citations
2. Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2arXiv:2408.05147Tom Lieberum, Senthooran Rajamanoharan, Arthur Conmy, et al.2024 · 115 citations
3. Attribution Patching Outperforms Automated Circuit DiscoveryarXiv:2310.10348Aaquib Syed, Can Rager, Arthur Conmy2023 · 81 citations
4. Chain-of-Thought Reasoning In The Wild Is Not Always Faithful ↗arXiv:2503.08679 · no summary yetIvan Arcuschin et al.2025 · 52 citations
5. Thought Anchors: Which LLM Reasoning Steps Matter? ↗arXiv:2506.19143 · no summary yetPaul C. Bogdan et al.2025 · 44 citations
6. Improving Dictionary Learning with Gated Sparse AutoencodersarXiv:2404.16014Senthooran Rajamanoharan, Arthur Conmy, Lewis Smith, et al.2024 · 42 citations
7. Successor Heads: Recurring, Interpretable Attention Heads In The Wild ↗arXiv:2312.09230 · no summary yetRhys Gould et al.2023 · 39 citations
8. Copy Suppression: Comprehensively Understanding an Attention Head ↗arXiv:2310.04625 · no summary yetCallum McDougall et al.2023 · 25 citations
9. AuditBench: Evaluating Alignment Auditing Techniques on Models with Hidden Behaviors ↗arXiv (Cornell University) · no summary yetAbhay Sheshadri et al.2026 · 6 citations

🤝 Co-authors here

10 co-authors with a page, the most papers together first.

• Janos Kramar ×2
• Lewis Smith ×2
• Neel Nanda ×2
• Rohin Shah ×2
• Senthooran Rajamanoharan ×2
• Tom Lieberum ×2
• Vikrant Varma ×2
• Aaquib Syed
• Anca Dragan
• Can Rager

🔗 Cited by

Papers of our library cite their works 697 times (of the 149,439 papers whose bibliographies we read); 132 citing papers with a page here →.

Updated 2 November 2026

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/authors/arthur-conmy/