PhD student, EPFL - EPF Lausanne
1 paper at NeurIPS 2025
The paper proves that a two-layer, single-head transformer can reliably perform in-context learning on any-order Markov chains.
