Visual Reasoning on HR-Bench 4K FCP
63.3AccuracyGPT-4o
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GPT-4o2026.02 | 63.3 | — | |
| Thyme2026.02 | 63 | — | |
| RTWIBackbone=DeepEyes-7B2026.02 | 60 | 53.4 | |
| SCBackbone=DeepEyes-7B2026.02 | 58 | — | |
| ASCBackbone=DeepEyes-7B2026.02 | 58 | 48.9 | |
| ESCBackbone=DeepEyes-7B2026.02 | 58 | 31.3 | |
| DeepconfBackbone=DeepEyes-7B2026.02 | 58 | 48 | |
| Self-Uncer.Backbone=DeepEyes-7B2026.02 | 58 | 40.2 | |
| CISCBackbone=DeepEyes-7B2026.02 | 56.8 | 37.1 | |
| Base*Backbone=DeepEyes-7B2026.02 | 56.5 | — |