Abstract page for arXiv paper 2610.10630: Phase-HDC: Replacing Optimizer History with Gradient Thresholds in Discrete Phase Learning