Cell-Level Visual Question Answering on PBSBench Out of Domain
97Accuracy (T/F)Quilt-LLaVA
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Quilt-LLaVAModel Category=Domain-specific - Pathology VL Model2026.04 | 97 | 39 | 4 | 11 | 9 | 13 | 58 | |
| PBS-VLModel Category=Ours2026.04 | 97 | 96 | 24 | 34 | 35 | 38 | 70 | |
| PathGen-LLaVAModel Category=Domain-specific - Pathology VL Model2026.04 | 96 | 78 | 8 | 19 | 14 | 19 | 59 | |
| DeepSeek-VL-7BModel Category=Open-source General-purpose VL Model2026.04 | 94 | 59 | 0 | 7 | 3 | 4 | 51 | |
| GPT-4oModel Category=Proprietary VL Models2026.04 | 86 | 93 | 9 | 28 | 10 | 15 | 65 | |
| Qwen2.5-VL-7BModel Category=Open-source General-purpose VL Model2026.04 | 77 | 84 | 1 | 10 | 4 | 7 | 58 | |
| LLaVA-1.6-Mistral-7BModel Category=Open-source General-purpose VL Model2026.04 | 74 | 58 | 2 | 9 | 5 | 8 | 55 | |
| LLaVA-1.6-Vicuna-7BModel Category=Open-source General-purpose VL Model2026.04 | 70 | 55 | 3 | 8 | 4 | 7 | 54 | |
| LLaVA-MedModel Category=Domain-specific - Medical VL Model2026.04 | 70 | 26 | 0 | 6 | 12 | 16 | 56 | |
| Qwen2-VL-7BModel Category=Open-source General-purpose VL Model2026.04 | 69 | 87 | 5 | 12 | 11 | 16 | 60 | |
| InternVL-8BModel Category=Open-source General-purpose VL Model2026.04 | 65 | 91 | 5 | 19 | 16 | 20 | 62 | |
| Gemini-1.5-proModel Category=Proprietary VL Models2026.04 | 64 | 92 | 2 | 24 | 6 | 10 | 56 | |
| Idefics3-8BModel Category=Open-source General-purpose VL Model2026.04 | 56 | 44 | 3 | 12 | 14 | 18 | 56 | |
| Claude-3.5-SonnetModel Category=Proprietary VL Models2026.04 | 54 | 90 | 1 | 28 | 5 | 10 | 60 | |
| BLIP2-OPT-6.7BModel Category=Open-source General-purpose VL Model2026.04 | 43 | 22 | 1 | 21 | 7 | 9 | 36 | |
| Med-flamingoModel Category=Domain-specific - Medical VL Model2026.04 | 25 | 26 | 0 | 4 | 7 | 8 | 38 | |
| PA-LLaVAModel Category=Domain-specific - Pathology VL Model2026.04 | 2 | 0 | 0 | 0 | 0 | 0 | 1 |