Skip to yearly menu bar Skip to main content


Oral Thu, Dec 10, 2026 • 3:45 PM – 4:00 PM AEDT Grand Ballroom B2-B3

Leaderboard Hacking: Preference-Based Model Evaluations are Vulnerable to Manipulation

Eve Fleisig ⋅ Joachim Baumann ⋅ Dylan Hadfield-Menell ⋅ Dirk Hovy

Abstract

Log in and register to view live content