Statistics as a tool for evaluating, auditing, and deploying black-box models and LLMs
Statistics has historically been the tool of choice for understanding and mitigating the operational risks of engineering deployments.
We need new statistical tools for the era of black-box models where the standard statistical ideas don't apply.
Does your work intersect with any of the following topics as they relate to LLMs and foundation models?
Benchmarks
Measuring and correcting bias
Automatic evaluation
Watermarking
Conformal prediction and other black-box uncertainty quantification techniques
Privacy
Auditing, safety, and risk analysis
If so, then this workshop is for you! See the Call for Papers for more details and submission instructions.
Mihaela
van der Schaar
Cambridge
Bernhard Schoelkopf
Max Planck Institute for Intelligent Systems
Virginia Smith
Carnegie Mellon
Weijie Su
University of Pennsylvania
Drew Prinster
Johns Hopkins
Sophia Sun
UC San Diego
Eleni Straitouri
Max Planck Institute