arXiv · NLP· Ang Li, Sean McLeish, Haozhe Chen, Nimit Kalra, Zaiqian Chen, Artem Gazizov, Venkata Anoop Suhas Kumar Morisetty, Bhavya Kailkhura, Harshitha Menon, Zhuang Liu, Brian R. Bartoldson, Tom Goldstein, Sanae Lotfi, Micah Goldblum, Pavel Izmailov·· 6 小时前AI 评分33
LCLM：通过 Encoder-Decoder 架构实现大规模端到端上下文压缩
End-to-End Context Compression at Scale
AI 导读
提出 LCLM（潜在上下文语言模型）压缩器，基于 0.6B-encoder 与 4B-decoder 架构，将 400B tokens 预训练数据压缩至 1:16 比率，显著提升推理效率。
来源：arXiv · NLP · arxiv.org