English→X machine translations generated with vLLM across 49 open-weight LLMs on three evaluation benchmarks. This dataset holds the raw model outputs (one CSV per model × dataset × target language); metric scores (BLEU / chrF / COMET / MetricX) are computed separately.

Layout

flores-200/<model>/eng-<lang>.csv   # 119 target languages
ntrex/<model>/eng-<lang>.csv        #  87 target languages
wmt24/<model>/eng-<lang>.csv        #  51 target languages

Each CSV has three columns:

column	description
`source`	English source sentence
`target`	reference translation from the benchmark
`translation`	model-generated translation

Source benchmarks

FLORES-200 (facebook/flores, devtest split)
NTREX (mteb/NTREX, test split)
WMT24++ (google/wmt24pp, train split)

Models (49)

aya-expanse-8b, aya-expanse-32b, tiny-aya-base, c4ai-command-r-08-2024, DeepSeek-R1-0528-Qwen3-8B, DeepSeek-R1-Distill-Llama-8B, DeepSeek-R1-Distill-Qwen-1.5B, DeepSeek-R1-Distill-Qwen-7B, DeepSeek-R1-Distill-Qwen-14B, DeepSeek-R1-Distill-Qwen-32B, Falcon3-3B-Instruct, Falcon3-7B-Instruct, Falcon3-10B-Instruct, gemma-3-1b-it, gemma-3-4b-it, gemma-3-12b-it, gemma-3-27b-it, gemma-4-E2B-it, gemma-4-E4B-it, gemma-4-26B-A4B-it, gemma-4-31B-it, glm-4-9b-chat, granite-4.1-3b, granite-4.1-8b, granite-4.1-30b, internlm2_5-7b-chat, internlm2_5-20b-chat, Llama-3.1-8B-Instruct, Llama-3.2-1B-Instruct, Llama-3.2-3B-Instruct, Ministral-3-3B-Instruct-2512, Ministral-3-8B-Instruct-2512, Ministral-3-14B-Instruct-2512, Mistral-7B-Instruct-v0.3, Mistral-Nemo-Instruct-2407, Mixtral-8x7B-Instruct-v0.1, Olmo-3-7B-Instruct, Olmo-3-1125-32B, Olmo-3.1-32B-Instruct, phi-4, Qwen3-1.7B, Qwen3-4B, Qwen3-8B, Qwen3-14B, Qwen3-30B-A3B, Qwen3.6-27B, Qwen3.6-35B-A3B, Yi-1.5-9B-Chat, Yi-1.5-34B-Chat.

Access

This is a gated dataset — access is reviewed and approved manually. Fill in the request form to be granted access.

Note: set the license field above to the license you intend to release under.

Downloads last month: -

Total file size:

15.1 GB