Papers and Projects - Josh Engels

Papers

How Transparent is DiffusionGemma? Joshua Engels, Callum McDougall, Bilal Chughtai, ..., Rohin Shah, and Neel Nanda. arXiv preprint, 2026. Paper | Blog | Twitter

Building Production-Ready Probes for Gemini. Janos Kramar, Joshua Engels, Zhengxuan Wang, Bilal Chughtai, Rohin Shah, Neel Nanda, and Arthur Conmy. arXiv preprint, 2026. Paper | Twitter

Training on Documents About Monitoring Leads to CoT Obfuscation. Reilly Haskins, Bilal Chughtai, and Joshua Engels. arXiv preprint, 2026. Paper | Blog | Twitter

When Reading the Chain of Thought Falls Short: A Testbed for Reasoning Trace Analysis. Daria Ivanova, Riya Tyagi, Joshua Engels, and Neel Nanda. Mechanistic Interpretability Workshop at ICML 2026. Paper | Blog

Designing Effective Monitor-Based Interventions for Mitigating Reward Hacking During RL. Aria Wong, Joshua Engels, and Neel Nanda. Mechanistic Interpretability Workshop at ICML 2026. Paper | Blog

Scaling Laws For Scalable Oversight. Joshua Engels*, David Baek*, Subhash Kantamneni*, and Max Tegmark. Neurips 2025 (Spotlight). Paper | Code | Twitter

Are Sparse Autoencoders Useful? A Case Study in Sparse Probing. Subhash Kantamneni*, Joshua Engels*, Senthooran Rajamanoharan, Max Tegmark, and Neel Nanda. ICML 2025. Paper | Blog | Code | Twitter

Low Rank Adapting Models for Sparse Autoencoders. Mathew Chen*, Joshua Engels*, and Max Tegmark. ICML 2025. Paper | Code | Twitter

Simple Mechanistic Explanations for Out-Of-Context Reasoning. Atticus Wang*, Joshua Engels*, Oliver Clive-Griffin*, Senthooran Rajamanoharan, and Neel Nanda. ICML 2025 Workshop on Reliable and Responsible Foundation Models. Paper

Dense SAE Latents Are Features, Not Bugs. Xiaoqing Sun, Alessandro Stolfo, Joshua Engels, Ben Wu, Senthooran Rajamanoharan, Mrinmaya Sachan, and Max Tegmark. Neurips 2025. Paper

Decomposing the Dark Matter of Sparse Autoencoders. Joshua Engels, Logan Smith, and Max Tegmark. TMLR 2025. Paper | Code | Twitter

The Geometry of Concepts: Sparse Autoencoder Feature Structure. Yuxiao Li, Eric J. Michaud, David D. Baek, Joshua Engels, Xiaoqing Sun, and Max Tegmark. Entropy 2025. Paper

Efficient Dictionary Learning with Switch Sparse Autoencoders. Anish Mudide, Joshua Engels, Eric J Michaud, Max Tegmark, and Christian Schroeder de Witt. ICLR 2025. Paper | Code | Twitter

Not All Language Model Features Are Linear. Joshua Engels, Eric J. Michaud, Isaac Liao, Wes Gurnee, and Max Tegmark. ICLR 2025. Paper | Code | Twitter | Talk

* indicates equal contribution