1
BLISS: A Lightweight Bilevel Influence Scoring Method for Data Selection in Language Model Pretraining
轻量级双层影响评分方法，高效筛选预训练数据，显著提升语言模型性能。
arXiv:2510.06048v4 Announce Type: replace Abstract: Effective data selection is essential for pretraining large language models (LLMs), enhancing effi…