SmallToLarge (S2L): Scalable Data Selection for Fine-tuning Large Language Models by Summarizing Training Trajectories of Small Models Paper • 2403.07384 • Published Mar 12 • 1
