Publications and manuscripts
Six studies on one question: whether machine learning results mean what they claim. 1 published, 2 submitted, 1 under review, 1 preprint and 1 in preparation. Every status is stated exactly and never upgraded.
Showing all 6
Which Leaderboard Orderings Survive a Human Reader?
A 400-item human audit certifies 174 of 351 pairwise orderings among 27 models under one stated assumption, and none without it.
- Benchmark Evaluation
- NLP · LLMs
Higher AUC, Fewer Cases Flagged: Subgroup Fairness Metrics Can Reverse at the Decision Threshold in Federated Diabetes Screening
Federated training narrows the White-Black AUC gap while a race-blind screening policy flags fewer Black respondents.
- Fairness
- Healthcare AI
- Privacy · Federated
Confidently Wrong: Ranking Inversion in Cross-Network Denial-of-Service Detection
On a new network, DoS detectors do not decay toward chance: many pass through it and score attacks below benign traffic.
- Network Security
- Deployment Shift
Comprehensive Evaluation of Machine Learning for Type 2 Diabetes Risk Prediction: Large-Scale External Validation and Fairness Analysis
Internally validated diabetes models lose discrimination on 1.28M external records, and adults over 60 lose the most.
- Healthcare AI
- Fairness
- Deployment Shift
IndiaFinBench: An Evaluation Benchmark for Large Language Model Performance on Indian Financial Regulatory Text
A public benchmark of 406 expert-annotated questions on Indian financial regulation, with twelve LLMs evaluated zero-shot.
- Benchmark Evaluation
- NLP · LLMs
- Financial ML
Persistent Racial Disparities in U.S. Mortgage Approval: Evidence from 42 Million Applications, 2020-2024
The public-data Black-White mortgage approval gap at national scale: how large, where it lives, and how it responds to institutional boundaries.
- Fairness
- Causal Inference