Evidence-first benchmark contamination and dataset integrity auditor
-
Updated
Aug 24, 2026 - Python
Evidence-first benchmark contamination and dataset integrity auditor
An open-source AI training supply chain security platform for dataset integrity, poisoning detection, model behavior analysis, provenance and ML artifact lineage.
Privacy-first dataset contamination investigation platform for human-written datasets. Explainable evidence, deterministic analysis, and human-controlled review.
Knowledge distillation vs structured pruning for on-collar cattle behaviour classification on microcontrollers, compared at matched footprint against a same-size control, after a dataset integrity check. Reproducible TinyML code.
An offline-first computer vision integrity and security assurance platform for auditing datasets, model artifacts, and inference outputs across ML pipelines.
OCT-C8 near-duplicate leakage audit using pHash, SSIM, fixed-model scoring, and paired retraining with ResNet18 and EfficientNet-B0.
Deterministic integrity and release controls for private evaluation datasets
To associate your repository with the dataset-integrity topic, visit your repo's landing page and select "manage topics."