FractAL: Soft Strategy Selection for Batch-Mode Active Learning
October 6, 2026
FractAL enables active learning practitioners to use multiple acquisition strategies within a single batch, rather than committing to one strategy at the start. This reduces the exploration overhead typically required by bandit-based strategy selection in high-throughput experiments.
HOW THIS AFFECTS YOU
●
builderThis allows for more efficient dataset curation when running many labeling rounds with large batch sizes.
●
researcherYou can optimize data labeling efficiency in high-throughput settings by implementing soft strategy selection.