热点
"LLM预训练" 相关文章
Rethinking Toxic Data in LLM Pretraining: A Co-Design Approach for Improved Steerability and Detoxification
MarkTechPost@AI 2025-05-14T04:00:41.000000Z
ByteDance Introduces QuaDMix: A Unified AI Framework for Data Quality and Diversity in LLM Pretraining
MarkTechPost@AI 2025-04-27T06:20:36.000000Z
比知识蒸馏好用，田渊栋等提出连续概念混合，再度革新Transformer预训练框架
机器之心 2025-02-16T08:07:41.000000Z
Instruction Pretraining LLMs
Ahead of AI 2024-10-22T06:07:39.000000Z