{"id":"arxiv/2607.01478","url":"https://arxiv.org/abs/2607.01478","tags":["math.OC","cs.CV","cs.LG"],"score":0,"title":"Boundary-Aware Quantization: Finite-Scale Decision Geometry of Neural Classifiers","domain":"math.OC","source":"arxiv","authors":["O. M. Kiselev"],"reviews":0,"subject":"math.OC","summary":"We measured quantization-induced decision-boundary changes using local logit-margin radii, first-order boundary displacement, normal variation, slice-boundary Jaccard distance, grid prediction changes, multiclass junction counts, and low-margin boundary-band flips. On the digits benchmark, 8-bit weight quantization preserved all test labels while producing boundary-mask Jaccard \\(0.428\\) on the PCA slice; at 4 bits, accuracy remained \\(0.9733\\), while boundary Jaccard rose to \\(0.970\\) and median local boundary shift reached \\(0.0290\\). Interpolation between adjacent quantization levels…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-07-01","reactions":{}}