[CL] Cite Pretrain: Retrieval-Free Knowledge Attribution for Large Language Models
[Duke University & Meta]
https://arxiv.org/abs/2506.17585