ensembles

This paper introduces the Conformal Relevance framework, which uses in-context learning and ensembling to improve conciseness in NLP tasks while maintaining coverage guarantees via conformal prediction.

DivSkill-SQL is a residual skill optimization framework that builds complementary agentic Text-to-SQL ensembles without model fine-tuning, improving selected accuracy by up to +11.1 points on Spider2-Lite by targeting examples that current ensembles fail on.