{"id":"arxiv/2610.27434","url":"https://arxiv.org/abs/2610.27434","tags":["stat.CO","math.NA","math.PR","math.ST"],"score":0,"title":"Beyond exponential tilting: sharp asymptotics and improved importance sampling for rare events","domain":"stat.CO","source":"arxiv","authors":["Mingxin Li","Hans Reimann","Timo Schorlepp","Georg Stadler"],"reviews":0,"subject":"stat.CO","summary":"A classical approach for importance sampling (IS) of a rare event is to tilt the measure using its large deviation minimizer, thereby shifting the proposal mean towards the event set. While this tilt is log-efficient, we show that the coefficient of variation of the resulting estimator still grows like $\\varepsilon^{-1/4}$ in the small-noise limit $\\varepsilon\\downarrow0$. Motivated by sharp large deviation asymptotics and the local geometry of the event set at the minimizer, we propose improvements using a control variate or an anisotropic, non-Gaussian proposal. We prove, up to truncations…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-10-20","reactions":{}}