Dataset · google
WMT24++: Translation into 55 Languages
google/wmt24pp
Human translations and post-editing for 55 en→xx pairs from the WMT24++ benchmark.
What's inside
The dataset contains human translations and post-editing data for 55 en→xx pairs released in WMT24++. Each language pair is stored in a separate JSONL file. This is an extension of WMT24 coverage to 55 languages and dialects; metrics and system outputs are separated into MTME.
What you can do with it
- Evaluate MT/LLM translators
- Fine-tune translation models on low-resource languages
- Analyze post-editing
- Tasks
- Translation
- Languages
- Arabic, Bulgarian, Bangla, Catalan, Danish, German, Greek, Spanish and 4 more
- Size
- 10K–100K rows
- Downloads this month
- 12.7K
- Likes
- 97
Models trained on it
- mradermacher/Seed-X-PPO-7B-GGUF 4.1K
- ByteDance-Seed/Seed-X-PPO-7BTranslation 943
- mradermacher/WiNGPT-Babel-2.1-i1-GGUF 484
- mradermacher/WiNGPT-Babel-2.1-GGUF 302
- mradermacher/WiNGPT-Babel-2.2-GGUF 197
- ByteDance-Seed/Seed-X-PPO-7B-AWQ-Int4Translation 164
- ByteDance-Seed/Seed-X-PPO-7B-GPTQ-Int8Translation 131
- winninghealth/WiNGPT-Babel-2-GGUFTranslation 104
How to load it
from datasets import load_dataset
ds = load_dataset("google/wmt24pp")