Composition Understanding on COMP-11 perception (test)
90.1RoT AccuracyCOMPASS
Evaluation Results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| COMPASSModel Type=unified2026.06 | 90.1 | 86.2 | 96.1 | 97.5 | 96.2 | 92.9 | 98.4 | 98.1 | 95.3 | 96.2 | 87.2 | 90.6 | |
| Qwen3-VLModel Type=understanding-only2026.06 | 56.7 | 70.8 | 78.7 | 76.9 | 54.1 | 54.6 | 87.2 | 86 | 61.8 | 72.5 | 66.2 | 66.4 | |
| BAGELModel Type=unified2026.06 | 51.1 | 66.4 | 80.2 | 68.2 | 54.2 | 57.9 | 87.4 | 83.2 | 70.7 | 74 | 52.8 | 63.2 | |
| InternVL3Model Type=understanding-only2026.06 | 50.6 | 65.7 | 57.5 | 55.9 | 55.8 | 64.3 | 82.8 | 78.3 | 65.5 | 62.3 | 52.6 | 60 | |
| Janus-ProModel Type=unified2026.06 | 50.2 | 55.1 | 56.7 | 66.8 | 52.7 | 51.5 | 71 | 69.5 | 55.5 | 51.5 | 52.3 | 54.2 | |
| AesExpertModel Type=understanding-only2026.06 | 42.3 | 50.6 | 53 | 50.2 | 47.8 | 50.4 | 63.9 | 56.3 | 54.4 | 58.1 | 50.1 | 49.5 | |
| Ming-Lite-UniModel Type=unified2026.06 | 31.4 | 68.1 | 67.4 | 63.2 | 63.9 | 57.3 | 84.2 | 82 | 69.8 | 65.8 | 62.2 | 58.2 |