Atlas · GenAI 2026
Data Drift
Production data distribution shifting from training data, degrading model performance
conceptPeak: 2021Monitoring & DriftAI consensus: 0/3
Prerequisites
No prerequisites.
Recommended reference
Rabanser et al. (2019) 'Failing Loudly: An Empirical Study of Methods for Detecting Dataset Shift'
Notes from AI deep research
Anthropic Opus
Models trained yesterday may fail today. Evidently AI, Great Expectations dla detection
Related skills
- → is part of: ML Monitoring(3/3)