{"id":"arxiv/2611.19163","url":"https://arxiv.org/abs/2611.19163","tags":["math.OC","cs.AI","cs.LG","eess.SY","stat.ML"],"score":0,"title":"Mean-field Token Steering","domain":"math.OC","source":"arxiv","authors":["Md Zulfiqur Haider","Asmaa Eldesoukey","Abhishek Halder","Yongxin Chen"],"reviews":0,"subject":"math.OC","summary":"We consider stochastic token steering through a deep stack of transformer layers as a mean-field control problem for an interacting particle system on the unit sphere, where interactions arise from the attention mechanism. In the mean-field limit, the token population evolves via a controlled integro-partial differential equation where the feedback control acts as a corrective vector field guiding the population toward an aggregate configuration of tokens in the representation space. We deduce the minimum effort control to accomplish this task over a fixed time horizon interpreted as…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-11-14","reactions":{}}