generalfactneutralTreating Bayesian update rules as experts and aggregating them can compete with the best expert in hindsight with bounded aggregation costMachine Learning (Statistics)28 Jul 2026http://arxiv.org/abs/2607.20239v1