Multimodal Search on VisBrowse (test)
34.7AccuracyVisual-Seeker
Evaluation Results
| Method | Links | |
|---|---|---|
| Visual-SeekerWorkflow Protocol=Multimodal Deep Search Agent2026.06 | 34.7 | |
| GPT-5Workflow Protocol=Agentic Workflow2026.06 | 29 | |
| Gemini-2.5-ProWorkflow Protocol=Agentic Workflow2026.06 | 26.6 | |
| GPT-5Workflow Protocol=Direct Answer2026.06 | 26 | |
| Gemini-2.5-FlashWorkflow Protocol=Agentic Workflow2026.06 | 20.7 | |
| Gemini-2.5-ProWorkflow Protocol=Direct Answer2026.06 | 19.5 | |
| Claude-4-SonnetWorkflow Protocol=Agentic Workflow2026.06 | 19.5 | |
| Gemini-2.5-FlashWorkflow Protocol=Direct Answer2026.06 | 16 | |
| Qwen3-VL-8B-InstructWorkflow Protocol=Agentic Workflow2026.06 | 15.4 | |
| Qwen3-VL-8B-InstructWorkflow Protocol=Direct Answer2026.06 | 10.7 | |
| Claude-4-SonnetWorkflow Protocol=Direct Answer2026.06 | 8.3 |