TY  - RPRT
TI  - When can Optimally Trained Transformers Achieve Near-Lossless Token Prediction for non-Markovian Data?
AU  - Kağan Akman
AU  - Jake Feldman Starosta
AU  - Naci Saldi
AU  - Serdar Yüksel
PY  - 2026
UR  - https://arxiv.org/abs/2610.41038
ID  - 2610.41038
ER  -