Reading Comprehension on Belebele Spanish (test)
91AccuracyTrinity Large
Evaluation Results
| Method | Links | |
|---|---|---|
| Trinity LargeArchitecture=MoE, Protocol=Zero-shot2026.02 | 91 | |
| Qwen3Parameters=8B, Protocol=Zero-shot2026.02 | 86 | |
| Qwen3Parameters=4B, Protocol=Zero-shot2026.02 | 85 | |
| DatologyParameters=8B, Protocol=Zero-shot2026.02 | 76 | |
| Llama-3.1Parameters=8B, Protocol=Zero-shot2026.02 | 74 | |
| Granite-4.0Parameters=Micro, Protocol=Zero-shot2026.02 | 74 | |
| DatologyParameters=3B, Protocol=Zero-shot2026.02 | 62 | |
| SmolLM3Parameters=3B, Protocol=Zero-shot2026.02 | 62 | |
| Llama-3.2Parameters=3B, Protocol=Zero-shot2026.02 | 56 | |
| LFM2.5Parameters=1.2B, Protocol=Zero-shot2026.02 | 54 | |
| Llama-3.2Parameters=1B, Protocol=Zero-shot2026.02 | 38 | |
| HQFiltering Strategy=Monolingual high-quality (HQ), Model size=1B2026.04 | 35.33 | |
| MLFiltering Strategy=Multilingual (ML) classifier, Model size=1B2026.04 | 34.56 | |
| No filteringFiltering Strategy=None, Model size=1B2026.04 | 34.22 |