Visual Reasoning on HR-Bench 8K FCP
59AccuracyRTWI
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RTWIBackbone=DeepEyes-7B2026.02 | 59 | 50.1 | |
| GPT-4o2026.02 | 58.5 | — | |
| DeepconfBackbone=DeepEyes-7B2026.02 | 58.3 | 43.9 | |
| Thyme2026.02 | 57.5 | — | |
| SCBackbone=DeepEyes-7B2026.02 | 57.3 | — | |
| ASCBackbone=DeepEyes-7B2026.02 | 57.3 | 47.4 | |
| ESCBackbone=DeepEyes-7B2026.02 | 57.3 | 24.7 | |
| CISCBackbone=DeepEyes-7B2026.02 | 56.8 | 34 | |
| Self-Uncer.Backbone=DeepEyes-7B2026.02 | 56.5 | 36.5 | |
| Base*Backbone=DeepEyes-7B2026.02 | 55.3 | — |