Scene Editing on E2A-Bench
69.1IF ScoreEdit-As-Act
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Edit-As-ActLatency (s)=87.2, Avg. Calls=5.9, Avg. Tok=190562026.03 | 69.1 | 86.6 | 91.7 | |
| SceneWeaverLatency (s)=102.5, Avg. Calls=7.3, Avg. Tok=183352026.03 | 68.7 | 78.3 | 82.1 | |
| Claude-4.5-opusLatency (s)=11.3, Avg. Calls=1.0, Avg. Tok=35142026.03 | 50.2 | 43.5 | 68.5 | |
| GPT-5Latency (s)=18.9, Avg. Calls=1.0, Avg. Tok=33572026.03 | 49.6 | 52.3 | 73.3 | |
| Gemini-3-Pro-previewLatency (s)=47.7, Avg. Calls=1.0, Avg. Tok=69592026.03 | 43.1 | 48.7 | 71.7 |