phenomenon · first seen 29 Oct 2026
transformer next-token prediction theory
Theoretical analysis of how transformers approximate optimal prediction for non-Markovian and hidden Markov sequence sources.
Mentions per week
PapersCodeDiscussion
Earliest sources
- When can Optimally Trained Transformers Achieve Near-Lossless Token Prediction for non-Markovian Data?Papers · 29 Oct 2026 · Kağan Akman, Jake Feldman Starosta, Naci Saldi, Serdar Yüksel
Latest sources
- When can Optimally Trained Transformers Achieve Near-Lossless Token Prediction for non-Markovian Data?Papers · 29 Oct 2026 · Kağan Akman, Jake Feldman Starosta, Naci Saldi, Serdar Yüksel