1 paper · 1 filter
Irene Lago, Ana Ezquerro, David Vilares
Interpreting properties of LLM-generated data is important for understanding its utility and limitations across learning tasks. In this work, we characterize synthetic data through…