Visual Reasoning on VisualProbe Easy
65.4AccuracyRTWI
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RTWIBackbone=Qwen3-VL Instruct2026.02 | 65.4 | 16.2 | |
| SCBackbone=Qwen3-VL Instruct2026.02 | 64 | — | |
| ASCBackbone=Qwen3-VL Instruct2026.02 | 64 | 20.7 | |
| ESCBackbone=Qwen3-VL Instruct2026.02 | 64 | 4.7 | |
| DeepconfBackbone=Qwen3-VL Instruct2026.02 | 61.9 | 14.6 | |
| RTWIBackbone=Qwen3-VL Thinking2026.02 | 60.4 | 44.3 | |
| DeepEyesBackbone=DeepEyes-7B2026.02 | 60.1 | — | |
| CISCBackbone=Qwen3-VL Instruct2026.02 | 59.7 | 12.8 | |
| Self-Cer.Backbone=Qwen3-VL Instruct2026.02 | 59 | 13.9 | |
| SCBackbone=Qwen3-VL Thinking2026.02 | 56.8 | — | |
| ASCBackbone=Qwen3-VL Thinking2026.02 | 56.8 | 37.2 | |
| ESCBackbone=Qwen3-VL Thinking2026.02 | 56.8 | 11.4 | |
| DeepconfBackbone=Qwen3-VL Thinking2026.02 | 56.8 | 42.9 | |
| BaseBackbone=Qwen3-VL Instruct2026.02 | 56 | — | |
| Self-Cer.Backbone=Qwen3-VL Thinking2026.02 | 54 | 23.9 | |
| CISCBackbone=Qwen3-VL Thinking2026.02 | 52.5 | 22.7 | |
| BaseBackbone=Qwen3-VL Thinking2026.02 | 47.6 | — | |
| GPT-4oBackbone=GPT-4o2026.02 | 47.5 | — |