LUKE MARKS
This is the personal website of Luke Marks.
I am reachable at marksluke076[at]gmail[dot]com.
Publications
- Interpreting learned feedback patterns in large language models (accepted to NeurIPS 2024)
- Informal safety guarantees for simulated optimizers through extrapolation from partial simulations
- Forthcoming - Mutual regularization for sparse autoencoders
Other
- Received a grant from Lightspeed Grants in 2023.
- Became a Non-Trivial Fellow in 2023.
- Was a research fellow at Apart Research in 2023.
- Became a Magnificent Grantee in 2024.
- Was a summer intern at the Torr Vision Group in 2024.
- Received a grant from BERI in 2024.
- Received a grant from Emergent Ventures in 2024.
- Dropped out of highschool in 2024.
Interests
- Sparse autoencoders for neural network interpretability
- Deep learning theory
- Mind uploading
- Homomorphic encryption
Projects
- BadTransformer: Simple transformer implemented with only NumPy and Python as dependencies.