{"id":"arxiv/2610.16819","url":"https://arxiv.org/abs/2610.16819","tags":["math.OC","cs.IT","cs.LG"],"score":0,"title":"Calibrated $α$-Barycenters for Multiobjective $f$-Divergence Optimization","domain":"math.OC","source":"arxiv","authors":["Dier Tang","Peijie Li","Xiaoyuan Zhang"],"reviews":0,"subject":"math.OC","summary":"Matching one distribution to many targets can require repeated evaluations of every target. CAL-$α$-MODM aggregates a nonlinear predictor into one barycenter, calibrates its residual derivatives, and checks steps against the original $f$-divergence objective. Under smoothness assumptions, calibration controls parameter-step model error without requiring targets to be close to the model. Exact-oracle stationarity and local convergence guarantees are established. At common matching, canonical geometry cancels the leading quadratic iteration error, giving cubic convergence with second-order…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-12","reactions":{}}