Video Question Answering on EgoCross (test)
79.2Animal AccuracyRD + timestamps
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| RD + timestampsModel=Gemini 3.1 Pro Preview2026.05 | 79.2 | 52.4 | 59.2 | 74.9 | 65.9 | 37.7 | 25.1 | 38.3 | |
| RDModel=Gemini 3.1 Pro Preview2026.05 | 71.6 | 50.4 | 53.5 | 74.9 | 62.5 | 37.7 | 25.2 | 38.27 | |
| RDModel=Gemini 3.1 Flash Image Preview2026.05 | 71 | 47.2 | 48.2 | 63.6 | 56.8 | 36.1 | — | 18.07 | |
| zero-shotModel=Gemini 3.1 Pro Preview2026.05 | 67.8 | 45.5 | 46.9 | 66.8 | 56.4 | 11.9 | — | 23.92 | |
| zero-shotModel=Gemini 3.1 Flash Image Preview2026.05 | 66.7 | 45.5 | 40 | 68.2 | 54.9 | 12.2 | — | 6.09 |