Multiple-choice Question Answering on NaturalBench
32.5Group AccuracyLlama-90B
Evaluation Results
| Method | Links | |
|---|---|---|
| Llama-90BModel=Llama-90B2026.06 | 32.5 | |
| OneVision-8BModel=OneVision-8B2026.06 | 28.7 | |
| OneVision-4B-LNModel=OneVision-4B-LN2026.06 | 27.8 | |
| OneVision-4BModel=OneVision-4B2026.06 | 27.6 | |
| InternVL2-8BModel=InternVL2-8B2026.06 | 24.4 | |
| Cambrian-8BModel=Cambrian-8B2026.06 | 20.3 | |
| LLaVA-13BModel=LLaVA-13B2026.06 | 14.7 | |
| LLaVA-7BModel=LLaVA-7B2026.06 | 14.2 |