AI, Data Science & Cloud Certificates from Google, IBM & Meta — 40% Off
One plan covers every Professional Certificate on Coursera. 40% off your first 3 months — limited time.
Unlock All Certificates
Dive into this 57-minute conference talk where Moritz Hardt from the Max Planck Institute for Intelligent Systems explores the evolving science of benchmarks in machine learning. Learn how benchmarks have been fundamental to research since the 1980s yet remain incompletely understood. Examine key insights about annotator errors, external validity of model rankings, and the potential of multi-task benchmarks that challenge traditional perspectives. This presentation from the 2024 SIAM Conference on Mathematics of Data Science offers valuable perspectives on advancing benchmark understanding in the machine learning community, covering topics like mathematical modeling, annotating practices, and data science applications.
