Visual Reasoning on Vstar Bench Spatial
81.6AccuracyRTWI
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RTWIBackbone=DeepEyes-7B2026.02 | 81.6 | 66.9 | |
| Thyme2026.02 | 80.3 | — | |
| Base*Backbone=DeepEyes-7B2026.02 | 80.3 | — | |
| SCBackbone=DeepEyes-7B2026.02 | 80.3 | — | |
| ASCBackbone=DeepEyes-7B2026.02 | 80.3 | 58.1 | |
| ESCBackbone=DeepEyes-7B2026.02 | 80.3 | 48.2 | |
| CISCBackbone=DeepEyes-7B2026.02 | 80.3 | 51.8 | |
| Self-Uncer.Backbone=DeepEyes-7B2026.02 | 80.3 | 56.8 | |
| DeepconfBackbone=DeepEyes-7B2026.02 | 79 | 56.8 | |
| GPT-4o2026.02 | 60.5 | — |