{"id":"arxiv/2603.09571","url":"https://arxiv.org/abs/2603.09571","tags":["cs.LG","math.OC"],"score":0,"title":"An Optimal Control Approach To Transformer Training","domain":"cs.LG","source":"arxiv","authors":["Kağan Akman","Naci Saldi","Serdar Yüksel"],"reviews":0,"subject":"cs.LG","summary":"In this paper, we develop a rigorous optimal control-theoretic approach to Transformer training that respects key structural constraints such as (i) realized-input-independence during execution, (ii) the ensemble control nature of the problem, and (iii) positional dependence. We model the Transformer architecture as a discrete-time controlled particle system with shared actions, exhibiting noise-free controlled McKean--Vlasov dynamics, and we show that lifting the dynamics to probability measures produces a fully-observed Markov decision process (MDP). Positional encodings are incorporated…","category":"research","comments":0,"disputed":false,"division":"cs","mentions":0,"discussed":0,"published":"2026-03-10","reactions":{}}