metadata
title: >-
Reproduction - Gradient Flow Sampler-based Distributionally Robust
Optimization
emoji: 🎯
colorFrom: blue
colorTo: indigo
sdk: static
pinned: false
tags:
- trackio
- trackio-logbook
- open-experiment
- icml2026-repro
- paper-QRtzkKrbJi
Gradient-flow DRO reproduction
A version-locked proof, rate, flow-time, half-bridge, and primary-experiment audit for all six registered claims.
python3 -m pip install -r requirements.txt
PYTHONWARNINGS=error python3 reproduce.py --output-dir replay-check
The three CIFAR-10 PGD plots are pinned primary v1 artifacts. The exact Gaussian and discrete calculations independently audit mechanisms and are not mislabeled as CIFAR reruns or universal proofs.