Visual Reasoning on VisualProbe Medium
40.6AccuracyDeepconf
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DeepconfBackbone=Qwen3-VL Instruct2026.02 | 40.6 | 16.7 | |
| RTWIBackbone=Qwen3-VL Instruct2026.02 | 40.6 | 18.5 | |
| SCBackbone=Qwen3-VL Instruct2026.02 | 40.2 | — | |
| ASCBackbone=Qwen3-VL Instruct2026.02 | 40.2 | 21.8 | |
| ESCBackbone=Qwen3-VL Instruct2026.02 | 40.2 | 8.8 | |
| Self-Cer.Backbone=Qwen3-VL Instruct2026.02 | 39.4 | 14.5 | |
| CISCBackbone=Qwen3-VL Instruct2026.02 | 38.6 | 14.7 | |
| BaseBackbone=Qwen3-VL Instruct2026.02 | 35.4 | — | |
| RTWIBackbone=Qwen3-VL Thinking2026.02 | 34.1 | 41.9 | |
| DeepEyesBackbone=DeepEyes-7B2026.02 | 29.8 | — | |
| SCBackbone=Qwen3-VL Thinking2026.02 | 29.6 | — | |
| ASCBackbone=Qwen3-VL Thinking2026.02 | 29.6 | 32.6 | |
| ESCBackbone=Qwen3-VL Thinking2026.02 | 29.6 | 10.9 | |
| DeepconfBackbone=Qwen3-VL Thinking2026.02 | 29.6 | 41 | |
| CISCBackbone=Qwen3-VL Thinking2026.02 | 27.7 | 20.4 | |
| Self-Cer.Backbone=Qwen3-VL Thinking2026.02 | 27.3 | 21.4 | |
| BaseBackbone=Qwen3-VL Thinking2026.02 | 25.8 | — | |
| GPT-4oBackbone=GPT-4o2026.02 | 15.4 | — |