Tag
Researchers introduce "Training Witnesses," a system for generating "honesty" certificates of ML training data and evaluations that conference reviewers can cheaply verify and trust, aiming to fix the broken trust model in ML conferences.