Named Entity Recognition on MasakhaNER Setswana 2.0
89.4Macro-F1 ScoreAfroXLMR-large
Evaluation results
| Method | Links | |
|---|---|---|
| AfroXLMR-largeSize=0.550B, Type=Encoder-Only2026.03 | 89.4 | |
| AfroXLMR-baseSize=0.270B, Type=Encoder-Only2026.03 | 87.7 | |
| AfriBERTaSize=0.126B, Type=Encoder-Only2026.03 | 83.2 | |
| MzansiLMVariant=mono-masakhaner-ft, Size=0.125B, Evaluation Protocol=Monolingual task-specific finetuning2026.03 | 38.9 | |
| MzansiLMVariant=multi-masakhaner-ft, Size=0.125B, Evaluation Protocol=Multilingual task-specific finetuning2026.03 | 23.7 | |
| Llama-3.1-70B-InstructSize=70B, Type=Decoder-Only2026.03 | 20.7 | |
| MzansiLMVariant=general-ft, Size=0.125B, Evaluation Protocol=Multi-task instruction finetuning2026.03 | 12.3 | |
| AfroLlama-V1Size=8B, Type=Decoder-Only2026.03 | 3.4 | |
| MzansiLMVariant=Base 0-shot, Size=0.125B, Evaluation Protocol=no finetuning and no in-context examples2026.03 | 0 | |
| InkubaLM-0.4BSize=0.4B, Type=Decoder-Only2026.03 | 0 | |
| Aya-101Size=13B, Type=Encoder-Decoder2026.03 | 0 |