Long-Context Aware Upcycling: A New Frontier for Hybrid LLM Scaling

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Fashi, Parsa Ashrafi, Saxena, Utkarsh, Rezagholizadeh, Mehdi, Jafari, Aref, Haridas, Akash, Yang, Mingyu, Bhatia, Vansh, Li, Guihong, Appia, Vikram, Barsoum, Emad
Format:	Preprint
Published:	2026
Subjects:	Computation and Language Machine Learning
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Internet

https://arxiv.org/abs/2604.24715