Saved in:
Bibliographic Details
Main Authors: Han, Sungjun, Suk, Juyoung, An, Suyeong, Kim, Hyungguk, Kim, Kyuseok, Yang, Wonsuk, Choi, Seungtaek, Shin, Jamin
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2504.15431
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910916202725376
author Han, Sungjun
Suk, Juyoung
An, Suyeong
Kim, Hyungguk
Kim, Kyuseok
Yang, Wonsuk
Choi, Seungtaek
Shin, Jamin
author_facet Han, Sungjun
Suk, Juyoung
An, Suyeong
Kim, Hyungguk
Kim, Kyuseok
Yang, Wonsuk
Choi, Seungtaek
Shin, Jamin
contents We introduce Trillion-7B, the most token-efficient Korean-centric multilingual LLM available. Our novel Cross-lingual Document Attention (XLDA) mechanism enables highly efficient and effective knowledge transfer from English to target languages like Korean and Japanese. Combined with optimized data mixtures, language-specific filtering, and tailored tokenizer construction, Trillion-7B achieves competitive performance while dedicating only 10\% of its 2T training tokens to multilingual data and requiring just 59.4K H100 GPU hours (\$148K) for full training. Comprehensive evaluations across 27 benchmarks in four languages demonstrate Trillion-7B's robust multilingual performance and exceptional cross-lingual consistency.
format Preprint
id arxiv_https___arxiv_org_abs_2504_15431
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Trillion 7B Technical Report
Han, Sungjun
Suk, Juyoung
An, Suyeong
Kim, Hyungguk
Kim, Kyuseok
Yang, Wonsuk
Choi, Seungtaek
Shin, Jamin
Computation and Language
Artificial Intelligence
Machine Learning
We introduce Trillion-7B, the most token-efficient Korean-centric multilingual LLM available. Our novel Cross-lingual Document Attention (XLDA) mechanism enables highly efficient and effective knowledge transfer from English to target languages like Korean and Japanese. Combined with optimized data mixtures, language-specific filtering, and tailored tokenizer construction, Trillion-7B achieves competitive performance while dedicating only 10\% of its 2T training tokens to multilingual data and requiring just 59.4K H100 GPU hours (\$148K) for full training. Comprehensive evaluations across 27 benchmarks in four languages demonstrate Trillion-7B's robust multilingual performance and exceptional cross-lingual consistency.
title Trillion 7B Technical Report
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2504.15431