FLORES+ is a multilingual machine translation evaluation benchmark covering over 200 language varieties. It is managed by the Open Language Data Initiative.
DatasetFree
Step 1: Log in to Hugging Face and accept the terms of use on the dataset page.
Step 2: Create an access token and log in with huggingface_hub.login().
Step 3: Install the datasets package with pip install datasets.
Step 4: Load all languages with load_dataset("openlanguagedata/flores_plus"), or pass a language code such as fra_Latn for one variety.
Step 5: Use the dev and devtest splits to evaluate translation systems, not to train them.
No. The card says it is for evaluation and should not be used as training data.
CC BY-SA 4.0. Access also requires accepting the conditions on Hugging Face.
The Open Language Data Initiative. The data was originally released by Meta researchers under the name FLORES.
The card asks you not to redistribute saved files publicly unless you protect them from automatic scraping.
MediaWiki MCPMediaWiki MCP Server lets LLM clients read and edit any MediaWiki wiki through MCP tools. It is made by Professional Wiki.
GDPvalGDPval is OpenAI's dataset of 220 real-world knowledge tasks across 44 occupations. It supports evaluating AI models on economically valuable work.