Multi-image reasoning and general capability evaluation on NLVR2
90.42AccuracyInternVL2.5
Evaluation results
| Method | Links | |
|---|---|---|
| InternVL2.5Params=8B2026.03 | 90.42 | |
| InternVL2.5 + CAPLParams=8B2026.03 | 90.13 | |
| Qwen2VLParams=7B2026.03 | 87.41 | |
| LLaVA-OVParams=7B2026.03 | 86.82 | |
| Idefics3Params=8B2026.03 | 85.14 | |
| GLM4.1VBaseParams=9B2026.03 | 84.98 | |
| GLM4.1VBase + CAPLParams=9B2026.03 | 84.87 | |
| Qwen2.5-VL + CAPLParams=7B2026.03 | 80.05 | |
| Qwen2.5-VLParams=7B2026.03 | 79.85 | |
| InternVL2Params=7B2026.03 | 77.68 | |
| Idefics2Params=8B2026.03 | 56.81 | |
| LLaVA-NextParams=7B2026.03 | 50.34 |