LLaMAX commited on
Commit
2536ecd
1 Parent(s): b4fac5f

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +19 -5
README.md CHANGED
@@ -183,10 +183,24 @@ We implement multiple versions of the LLaMAX model, the model links are as follo
183
  If our model helps your work, please cite this paper:
184
 
185
  ```
186
- @article{lu2024llamax,
187
- title={LLaMAX: Scaling Linguistic Horizons of LLM by Enhancing Translation Capabilities Beyond 100 Languages},
188
- author={Lu, Yinquan and Zhu, Wenhao and Li, Lei and Qiao, Yu and Yuan, Fei},
189
- journal={arXiv preprint arXiv:2407.05975},
190
- year={2024}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
191
  }
192
  ```
 
183
  If our model helps your work, please cite this paper:
184
 
185
  ```
186
+ @inproceedings{lu-etal-2024-llamax,
187
+ title = "{LL}a{MAX}: Scaling Linguistic Horizons of {LLM} by Enhancing Translation Capabilities Beyond 100 Languages",
188
+ author = "Lu, Yinquan and
189
+ Zhu, Wenhao and
190
+ Li, Lei and
191
+ Qiao, Yu and
192
+ Yuan, Fei",
193
+ editor = "Al-Onaizan, Yaser and
194
+ Bansal, Mohit and
195
+ Chen, Yun-Nung",
196
+ booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2024",
197
+ month = nov,
198
+ year = "2024",
199
+ address = "Miami, Florida, USA",
200
+ publisher = "Association for Computational Linguistics",
201
+ url = "https://aclanthology.org/2024.findings-emnlp.631",
202
+ doi = "10.18653/v1/2024.findings-emnlp.631",
203
+ pages = "10748--10772",
204
+ abstract = "Large Language Models (LLMs) demonstrate remarkable translation capabilities in high-resource language tasks, yet their performance in low-resource languages is hindered by insufficient multilingual data during pre-training. To address this, we conduct extensive multilingual continual pre-training on the LLaMA series models, enabling translation support across more than 100 languages. Through a comprehensive analysis of training strategies, such as vocabulary expansion and data augmentation, we develop LLaMAX. Remarkably, without sacrificing its generalization ability, LLaMAX achieves significantly higher translation performance compared to existing open-source LLMs (by more than 10 spBLEU points) and performs on-par with specialized translation model (M2M-100-12B) on the Flores-101 benchmark. Extensive experiments indicate that LLaMAX can serve as a robust multilingual foundation model. The code and the models are publicly available.",
205
  }
206
  ```