X-EcoMLA: Upcycling Pre-Trained Attention into MLA for Efficient and Extreme KV Compression

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Li, Guihong, Rezagholizadeh, Mehdi, Yang, Mingyu, Appia, Vikram, Barsoum, Emad
Format:	Preprint
Published:	2025
Subjects:	Computation and Language
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Internet

https://arxiv.org/abs/2503.11132