Multimodal Multi-choice on BLINK
60AccuracyGPT-4o
Evaluation results
| Method | Links | |
|---|---|---|
| GPT-4o2025.12 | 60 | |
| GPT-4v2025.12 | 58.3 | |
| SkiLa-V2025.12 | 56.7 | |
| SkiLa2025.12 | 56.7 | |
| Claude3.7-Sonnet2025.12 | 56.6 | |
| Direct SFT2025.12 | 56.1 | |
| Qwen2.5-VL 7BModel Size=7B2025.12 | 55.1 | |
| GPT-4o-mini2025.12 | 53.6 | |
| Gemma3 27BModel Size=27B2025.12 | 53.5 | |
| LLaVA-OneVision 7BModel Size=7B2025.12 | 53 | |
| LVR 7B*Model Size=7B, Tested by authors=true2025.12 | 52.5 | |
| Vision-R1 7B*Model Size=7B, Tested by authors=true2025.12 | 51 | |
| ROSS 7BModel Size=7B2025.12 | 45.3 | |
| Cambrian 13BModel Size=13B2025.12 | 43.1 | |
| Janus-Pro 7BModel Size=7B2025.12 | 38.7 |