Announcement_6

Selective prediction allows AI systems to abstain when outputs are unreliable. How can we move beyond binary risks while retaining finite-sample validity? Check out our new paper SCoRE, where we develop an e-value-based framework for general risk control in selective prediction.