Alexandria - Human Dialectness Rating (Mauritanian Arabic): leaderboard
Human evaluation of English to Mauritanian Arabic translation on the Alexandria (El Mekki et al., ACL 2026) test set: native speakers rate on a 1-5 scale whether each translation sounds like a native speaker of the target dialect rather than Modern Standard Arabic or a mix, independently of meaning. Mean rating, higher is better.
Metric: Dialectness and fluency (1-5; native Mauritanian speakers rate whether the English-to-dialect translation sounds like a native speaker of the target dialect). Source: arxiv.org. Saturation forecast: Around October 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) (Minimal) | 3.04 | #141 (Gemini 3 Flash (Preview)) |
| 2 | Gemini 3 Flash (Preview) (Medium) | 3.01 | #141 (Gemini 3 Flash (Preview)) |
| 3 | c4ai-command-a-03-2025 | 2.27 | #1096 |
| 4 | GPT-OSS-120B | 2.13 | #699 |
| 5 | Gemma 3 27B (IT) | 2.08 | #1306 |
| 6 | aya-expanse-32B | 2.04 | #1950 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=alexandria-human-dialectness-rating-mauritanian-arabic · How It Works · Data refreshed daily, snapshot 2026-11-12.