Language Identification on Amazon Massive
99.67In-domain AccuracySonnet 3.5.
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Sonnet 3.5.prompting=strong tailored prompt2025.12 | 99.67 | 97.9 | 97.77 | 97.77 | — | 97.79 | |
| Baseline2025.12 | 99.58 | 99.2 | 99.2 | 99.2 | 99.52 | 99.2 | |
| PolyLinguatraining=end-to-end2025.12 | 99.51 | 99.25 | 99.25 | 99.25 | 99.49 | 99.25 | |
| Baseline+SupConloss=supervised contrastive learning2025.12 | 99.27 | 99.24 | 99.23 | 99.23 | 99.47 | 99.23 | |
| FastText2025.12 | 97.74 | 94.56 | 92.95 | 92.95 | — | 93.2 | |
| Langdetect2025.12 | 95.08 | 79.17 | 79.29 | 79.29 | — | 78.6 | |
| XLM-LID2025.12 | 82.59 | 97.43 | 96.02 | 96.02 | 98.64 | 96.65 |