Visual Grounding on V* Direct Attributes 52
90.43AccuracyDeepEyes
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DeepEyesBase Model=DeepEyes2026.03 | 90.43 | 1 | |
| SpecEyesBase Model=DeepEyes, Aggregation Strategy=min2026.03 | 90.43 | 1.53 | |
| SpecReasonBase Model=Thyme2026.03 | 89.57 | 0.48 | |
| SpecEyesBase Model=Thyme, Aggregation Strategy=min2026.03 | 87.83 | 1.32 | |
| ThymeBase Model=Thyme2026.03 | 86.96 | 1 | |
| SpecEyesBase Model=DeepEyes, Aggregation Strategy=log2026.03 | 83.48 | 2.06 | |
| SpecEyesBase Model=DeepEyes, Aggregation Strategy=bottom2026.03 | 83.48 | 2.13 | |
| SpecEyesBase Model=Thyme, Aggregation Strategy=log2026.03 | 80.87 | 1.82 | |
| SpecReasonBase Model=DeepEyes2026.03 | 80.19 | 0.61 | |
| SpecEyesBase Model=DeepEyes, Aggregation Strategy=mean2026.03 | 78.26 | 2.89 | |
| SpecEyesBase Model=Thyme, Aggregation Strategy=bottom2026.03 | 78.26 | 2.18 | |
| Qwen3-VL-2BMode=draft only2026.03 | 77.39 | 5.44 | |
| SpecEyesBase Model=Thyme, Aggregation Strategy=mean2026.03 | 77.39 | 2.34 |