Back to people
Mamoru B Komachi@mamoruk
Recent posts
Our paper accepted to EMNLP 2024 Findings is now available on arXiv🎉 The key finding is that pruning using few-shot translation demonstrations as calibration data enhances non-English performance in multilingual large language models like XGLM. Thank you to our collaborators. Please check it out👀 https://t.co/f5vXGfgdcC
M
EMNLP 2024 に @HwichanKim さんと @drJunSuzuki さんとの共同研究、Pruning Multilingual Large Language Models for Multilingual Inference が Findings で採択されました〜。よくよく調べると国内では発表してなかった（YANS 2022 は違う）ですが、多言語 LLM の転移能力の分析と改善のお話です。