Real-world QA on RealworldQA v1.0 (test)
75.5ScoreGPT-4o
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-4oTool Use=No, Param Size=-2026.03 | 75.5 | |
| Qwen2.5-VLTool Use=No, Param Size=32B2026.03 | 70.2 | |
| ThymeTool Use=Yes, Param Size=7B2026.03 | 70.2 | |
| InternVL2.5Tool Use=No, Param Size=8B2026.03 | 70.1 | |
| VRETool Use=No, Param Size=7B2026.03 | 69.8 | |
| Qwen2.5-VLTool Use=No, Param Size=7B2026.03 | 68.2 | |
| LLaVA-OneVisionTool Use=No, Param Size=7B2026.03 | 66.3 |