Salvatore Greco
@salvatoregreco.bsky.social
Postdoc Researcher at King's College London | HCI - Fairness - NLP | Prev: PhD at Politecnico di Torino - 2xVisiting @BellLabs
📊 Experiments across classifiers for text, image, and audio data show that DriftLens:
(1) outperforms prior methods in 15/17 cases
(2) runs ≥5× faster
(3) produces drift curves closely aligned with actual drift
(4) identifies representative drift samples as explanations (3/3)
(1) outperforms prior methods in 15/17 cases
(2) runs ≥5× faster
(3) produces drift curves closely aligned with actual drift
(4) identifies representative drift samples as explanations (3/3)
September 16, 2025 at 9:20 AM
📊 Experiments across classifiers for text, image, and audio data show that DriftLens:
(1) outperforms prior methods in 15/17 cases
(2) runs ≥5× faster
(3) produces drift curves closely aligned with actual drift
(4) identifies representative drift samples as explanations (3/3)
(1) outperforms prior methods in 15/17 cases
(2) runs ≥5× faster
(3) produces drift curves closely aligned with actual drift
(4) identifies representative drift samples as explanations (3/3)
✨ Key features:
(1) no labels required (completely unsupervised)
(2) real-time execution
(3) detects, characterizes & explains drift (2/3)
(1) no labels required (completely unsupervised)
(2) real-time execution
(3) detects, characterizes & explains drift (2/3)
September 16, 2025 at 9:20 AM
✨ Key features:
(1) no labels required (completely unsupervised)
(2) real-time execution
(3) detects, characterizes & explains drift (2/3)
(1) no labels required (completely unsupervised)
(2) real-time execution
(3) detects, characterizes & explains drift (2/3)