Unlabeled Data Can Provably Enhance In-Context Learning of Transformers
Published in NeurIPS 2025, 2025
This paper investigates the theoretical foundations of how unlabeled data in the prompt improves In-Context Learning.
Recommended citation: Renpu Liu, Jing Yang. (2025). "Unlabeled Data Can Provably Enhance In-Context Learning of Transformers." NeurIPS 2025.
Download Paper