Multi-image Understanding on Blink (test)
56.8AccuracyInternVL2-Llama3-76B
Evaluation results
| Method | Links | |
|---|---|---|
| InternVL2-Llama3-76B2026.01 | 56.8 | |
| GPT-4V2026.01 | 54.6 | |
| attention-masking strategybackbone=LLaVA-OV-7B, attention masking=true2026.01 | 51.9 | |
| LLaVA-OV-7B2026.01 | 50.4 | |
| LLaVA-OV-0.5B2026.01 | 40.4 | |
| attention-masking strategybackbone=LLaVA-OV-0.5B, attention masking=true2026.01 | 39.1 | |
| procedural data-generation strategybackbone=LLaVA-OV-0.5B, attention masking=false2026.01 | 38.9 | |
| LLaVA-v1.5-7B2026.01 | 37.1 | |
| InternVL2-8B2026.01 | 23.4 | |
| Qwen2VL-7B2026.01 | 17.7 | |
| InternVL2-2B2026.01 | 16.3 | |
| Qwen2VL-2B2026.01 | 12.7 |