Science Question Answering on ARC-C French (test)
90AccuracyTrinity Large (MoE)
Evaluation results
| Method | Links | |
|---|---|---|
| Trinity Large (MoE)Architecture=MoE, Evaluation=zero-shot2026.02 | 90 | |
| Qwen3 8BModel size=8B, Evaluation=zero-shot2026.02 | 87 | |
| Qwen3 4BModel size=4B, Evaluation=zero-shot2026.02 | 84 | |
| Datology 8BModel size=8B, Evaluation=zero-shot2026.02 | 74 | |
| Granite-4.0 MicroEvaluation=zero-shot2026.02 | 70 | |
| Llama-3.1 8BModel size=8B, Evaluation=zero-shot2026.02 | 68 | |
| SmolLM3 3BModel size=3B, Evaluation=zero-shot2026.02 | 68 | |
| LFM2.5 1.2BModel size=1.2B, Evaluation=zero-shot2026.02 | 65 | |
| Datology 3BModel size=3B, Evaluation=zero-shot2026.02 | 58 | |
| Llama-3.2 3BModel size=3B, Evaluation=zero-shot2026.02 | 54 | |
| Llama-3.2 1BModel size=1B, Evaluation=zero-shot2026.02 | 31 | |
| ArrowBase Model=Phi-2, Evaluation Setting=Zero-shot2025.05 | 27.51 | |
| Phi-2Base Model=Phi-2, Evaluation Setting=Zero-shot2025.05 | 26.84 | |
| GenKnowSubBase Model=Phi-2, Evaluation Setting=Zero-shot, Setting=En2025.05 | 25.5 | |
| GenKnowSubBase Model=Phi-2, Evaluation Setting=Zero-shot, Setting=Avg2025.05 | 24.16 |