Scholar
Leon Lang
Google Scholar ID: E3ae_sMAAAAJ
PhD Student, University of Amsterdam
Citations & Impact
Citations
407
H-index
5
i10-index
4
Publications
14
Co-authors
23
list available
Publications
Resume (English only)
Academic Achievements
- Published multiple papers on AI alignment, risks of optimizing learned reward functions, and partial observability challenges in RLHF.
- Proposed a general method to build E(N)-equivariant steerable CNNs based on the Wigner-Eckart theorem.
- Generalized Hu's Theorem for information decomposition beyond Shannon entropy, including Kolmogorov complexity and generalization error.
- Developed factored space models as a new foundation for causality across abstraction levels.
- Researched modeling human beliefs about AI behavior to improve scalable oversight.
