2 papers
stat.ML2026
Did We Actually Fix It? An Independent Adversarial Stress-Test of Post-Point-Adjustment Evaluation Metrics for Time-Series Anomaly Detection
Zongye Lyu
Point-adjustment (PA), for years the default scoring protocol in time-series anomaly detection (TSAD), was shown by Kim et al. (2022) to award near-perfect F1 to random anomaly sco…
cs.LG2026
Exact and Certified Data Shapley for Weighted k-Nearest-Neighbor Regression and Soft-Label Prediction
Zongye Lyu
Data Shapley answers which training points are worth what, and its nearest-neighbor specialization is the version actually deployed, shipped by toolkits such as pyDVL and OpenDataV…