{"id":"arxiv/2602.02005","url":"https://arxiv.org/abs/2602.02005","tags":["cs.AR","cs.LG","eess.SY","hep-ex","quant-ph"],"score":0,"title":"Position: The Need for Ultrafast Training","domain":"cs.AR","source":"arxiv","authors":["Duc Hoang"],"reviews":0,"subject":"cs.AR","summary":"Domain-specialized FPGAs have delivered unprecedented performance for low-latency inference across scientific and industrial workloads, yet nearly all existing accelerators assume static models trained offline, relegating learning and adaptation to slower CPUs or GPUs. This separation fundamentally limits systems that must operate in non-stationary, high-frequency environments, where model updates must occur at the timescale of the underlying physics. In this paper, I argue for a shift from inference-only accelerators to ultrafast on-chip learning, in which both inference and training…","category":"research","comments":0,"disputed":false,"division":"cs","mentions":0,"discussed":0,"published":"2026-02-02","reactions":{}}