Pillar 2 Intra-shot prompt-following alignment
96.7Intra-shot Character PresenceEntityMem
Evaluation results
| Method | Links | ||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| EntityMem2026.05 | 96.7 | 88.8 | 68.7 | 74 | 60.7 | 68.4 | 80.2 | 72.6 | 60.1 | 71.2 | 69.1 | 72.8 | 57.3 | 55.5 | 60.3 | 70.6 | 56.2 | 55.7 | 61.8 | 51.9 | 70.6 | 69.7 | 78.1 | 71.6 | |
| HoloCine2026.05 | 88.2 | 72.3 | 62.4 | 34.9 | 36.9 | 48.2 | 33.9 | 44.9 | 26.7 | 37.3 | 33.1 | 38.3 | 25.6 | 30.6 | 35.4 | 47.4 | 30.5 | 34.6 | 56.9 | 45.8 | 60.6 | 69.5 | 61.6 | 77.2 | |
| StoryMem2026.05 | 84.9 | 89.3 | 68.1 | 45.2 | 42.4 | 48.5 | 50.4 | 53.9 | 61.8 | 70.1 | 70.9 | 71.5 | 59.8 | 50.4 | 52.9 | 62.7 | 52.2 | 52 | 54.7 | 44.6 | 59.5 | 62.6 | 71.2 | 72.3 | |
| CineTrans2026.05 | 79.6 | 77.6 | 65.1 | 32.7 | 36.6 | 41.3 | 37.8 | 52.1 | 38.4 | 50.8 | 48 | 53.9 | 37.1 | 42.8 | 47.4 | 58.8 | 42.9 | 48.8 | 27.3 | 12.4 | 47.8 | 32.3 | 34.6 | 52.8 |