Code for the paper "ClipMind: A Framework for Auditing Short-Format Video Recommendations Using Multimodal AI Models"
-
Updated
Jun 26, 2025 - Jupyter Notebook
Code for the paper "ClipMind: A Framework for Auditing Short-Format Video Recommendations Using Multimodal AI Models"
Cryptographically signed audit evidence and verification for privacy-preserving computation, with CKKS and FHE integrations.
Persona-driven agents that build real signed-in YouTube watch histories, for auditing algorithmic personalization
The comprehensive list of AI governance tools: 130+ platforms and resources for AI risk, compliance, model governance, audit, fairness, privacy, and responsible AI.
Open-source statistical engine for auditing AI employment decision systems: adverse impact analysis under the Uniform Guidelines (29 CFR 1607), name-swap bias testing, and model drift detection
Pilot audit of cultural bias in vision–language models: LLaVA-NeXT associates South Asian subjects with domestic roles 9× more often, and exoticises South Asian cultural events 14× more.
Bias audit toolkit reproducing the recruitment AI case from Activity 6 — Unidad 3, Electiva II Inteligencia Artificial Avanzada. Empirical analysis with reweighing mitigation, Model Card documentation, and APA 7 references.
Does an LLM told to favour a demographic group say so in its written justification? Collection and analysis harness for a four-arm crossed study of disclosure in LLM résumé screening.
Systematic safety audit of a ResNet-50 diabetic retinopathy classifier using the Medical Algorithmic Audit framework (Liu et al., 2022). Included error analysis, subgroup testing, adversarial robustness, and FMEA risk scoring on the APTOS 2019 dataset.
Auditing a 0.928-AUC loan-approval model for fairness. Accuracy was never the problem: employment status carries the disparity, and a default 0.5 threshold quietly penalises students and unemployed borrowers. NYU Responsible Data Science.
Ahimsa Impact Assessment (AIA v1.1): Dual-axis evaluation of non-extraction (Himsa risk) and affirmative flourishing (Sarvodaya) with RFC 8785 LOKAHI manifest compilation, 50-system benchmark, and 20 historical disasters suite.
To associate your repository with the algorithmic-auditing topic, visit your repo's landing page and select "manage topics."