BCL: Bayesian In-Context Learning Framework for Information Extraction

In three linesBCL is an optimization framework for information extraction using particle filtering and Bayesian updates to systematically refine label representations. It generalizes across sequence labeling and relation classification tasks, demonstrating consistent improvements over existing approaches across model scales.

Summary generated by Claude — human-verified