bayesian-in-context-learning

BCL is the first optimization framework that uses particle filtering with Bayesian updates to systematically refine label representations for information extraction tasks, showing consistent improvements over existing methods.