Dataset · facebook
FLORES-200 machine translation benchmark
facebook/flores
Parallel texts for evaluating translation from English into multiple languages, including low-resource ones.
What's inside
FLORES is a benchmark for machine translation between English and low-resource languages. The FLORES-200 version doubles the language coverage of FLORES-101; metadata lists languages such as ace, acm, acq, aeb, af, ajp, and others. Volume ranges from 1M to 10M records in tabular/text format. The repository is no longer updated, with openlanguagedata/flores_plus serving as the new version.
What you can do with it
- Evaluating MT quality on low-resource languages
- Testing multilingual LLMs
- Comparing translation models on a standard benchmark
- Tasks
- Text generation, Translation
- Languages
- Acehnese, acm, acq, Tunisian Arabic, Afrikaans, apc, Akan, Albanian and 4 more
- Size
- 1M–10M rows
- Downloads this month
- 6.2K
- Likes
- 135
- Access
- on request on Hugging Face
Models trained on it
- mradermacher/Seed-X-PPO-7B-GGUF 4.1K
- mradermacher/Aitana-7B-S-Instruct-i1-GGUFText generation 2.8K
- ByteDance-Seed/Seed-X-Instruct-7BTranslation 2.1K
- gplsi/Aitana-7B-S-InstructText generation 1.2K
- gplsi/Aitana-2B-S-InstructText generation 976
- ByteDance-Seed/Seed-X-PPO-7BTranslation 943
- mradermacher/Aitana-7B-S-Instruct-v0.1-i1-GGUFText generation 770
- mradermacher/Seed-X-Instruct-7B-GGUF 556
How to load it
from datasets import load_dataset
ds = load_dataset("facebook/flores")