A Data-Efficient Path to Multilingual LLMs: Language Expansion via Post-training PARAMš„ Integration into Upcycled MoE
The method expands multilingual LLMs via post-training PARAMĪ integration into upcycled MoE for data-efficient language acquisition.
Published 2026Paper ā
67%
OverallHighly rated
?
OverallHighly ratedVote to see the scoreThe exact score shows once you've voted, so every vote is your own call. The first half of each home page shelf shows its scores.
Readers
ā
Only vote on papers you've read. Sign in with GitHub to vote.
AI panel2/20reviewers recommend it
lenient 1/5
medium 1/10
strict 0/5
AI panel?Vote to see what the 20 AI reviewers said
PARAMĪ offers a compelling, audit-worthy router map for multilingual MoE expansion, though its data-efficient gains depend on favorable topology and still lack rigorous dense-adapter comparisons.
Abstract
Hao Zhou, Tianhao Li, Zhijun Wang, Shuaijie She, Linjuan Wu, Hao-Ran Wei, Baosong Yang, Jiajun Chen, Shujian Huang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.