Title: Alex Makelov
Description: Personal website of Alex Makelov - researcher in mechanistic interpretability of large language models.
Domain: amakelov.github.io
Links:
| Prof. Aleksander Madry | https://madry.mit.edu/ |
| Prof. Salil Vadhan | https://salil.seas.harvard.edu/ |
| mandala | https://github.com/amakelov/mandala |
| Google Scholar | https://scholar.google.com/citations?user=haO4sKoAAAAJ&hl=en&oi=ao |
| Semantic Scholar | https://www.semanticscholar.org/author/Aleksandar-Makelov/17775913 |
| https://x.com/AMakelov | |
| Persona Features Control Emergent Misalignment | https://arxiv.org/abs/2506.19823 |
| Sparse Autoencoders Match Supervised Features for Model Steering on the IOI Task | https://openreview.net/attachment?id=JdrVuEQih5&name=pdf |
| AI Alignment Forum post | https://www.alignmentforum.org/posts/zj3GKWAnhPgTARByB/saes-discover-meaningful-features-in-the-ioi-task |
| Towards Principled Evaluations of Sparse Autoencoders for Interpretability and Control | https://arxiv.org/abs/2405.08366 |
| mandala: Compositional Memoization for Simple & Powerful Scientific Data Management | https://amakelov.github.io/scipy-mandala.pdf |
| Is This the Subspace You Are Looking for? An Interpretability Illusion for Subspace Activation Patching | https://arxiv.org/abs/2311.17030 |
| Backdoor or Feature? A New Perspective on Data Poisoning | https://openreview.net/pdf?id=4NT3umNU3D0 |
| Towards Deep Learning Models Resistant to Adversarial Attacks | https://arxiv.org/abs/1706.06083 |
| Expansion in Lifts of Graphs | https://dash.harvard.edu/bitstream/handle/1/14398532/MAKELOV-SENIORTHESIS-2015.pdf?sequence=1 |
| Practical dependency tracking for Python function calls | https://amakelov.github.io/blog/deps/ |
| Mandala: Python programs that save, query and version themselves | https://amakelov.github.io/blog/pl/ |
Viewport: width=device-width, initial-scale=1