Skip to content

Benchmark

Persian-family Machine Translation BLEU (Dari variant, source-pooled, zero-shot)

Multilingual benchmarks compare model performance across languages. Higher scores are better.

Models measured
2
Scores recorded
2
Sources
1
IQ weight
No weight published

Score distribution

13.4
Mean
16.4
Median
3.1
Std dev
10.3
10th percentile
16.4
90th percentile
10.3
Min
10.316.4
The band covers the middle 80% of scores. The line marks the median.

Leaderboard

#ModelScore
1Qwen2.5 3B Instructtable: 6, task: Machine Translation, metric: BLEU, subset: Dari variant, source-pooled· Persian, Dari and Tajiki paper data docs16.4
2Qwen2.5 1.5B Instructtable: 6, task: Machine Translation, metric: BLEU, subset: Dari variant, source-pooled· Persian, Dari and Tajiki paper data docs10.3

Data credits: Noor Mairukh Khan Arnob and Abu Bakar Siddique Mahi. One Language, Three of Its Voices, SilkRoadNLP 2026, ©2026 ACL, CC BY 4.0. Numeric/protocol extract from Tables 2, 5 and 6; language and pooled subsets retained. (licence)