metadata
title: Reproduction - High-Dimensional Learning Dynamics of Quantized Models with STE
emoji: 🔢
colorFrom: blue
colorTo: blue
sdk: static
pinned: false
tags:
- trackio
- trackio-logbook
- open-experiment
- icml2026-repro
- paper-bI9moH3UZw
Reproduction - High-Dimensional Learning Dynamics of Quantized Models with STE
Independent deterministic audit of all five registered claims using a 20,000-particle mean-field SDE, the source macroscopic ODE, 48 finite-dimension one-pass STE runs, plateau/drop sweeps, fixed-point stability controls, and all three small-step asymptotic regimes. All 15 scientific gates pass in two warning-strict byte-identical runs.
Primary source: https://arxiv.org/pdf/2510.10693v1 OpenReview: https://openreview.net/forum?id=bI9moH3UZw OpenReview paper ID: bI9moH3UZw arXiv version: 2510.10693v1 Primary PDF SHA-256: 6572a9be1af275679eeb66c0d47d147a0f88cbef92f2ff1529516d422d4fe6bc Implementation: independent from the formulas and algorithms in the pinned PDF. No competitor artifact or result was used.