Video Detailed Captioning on VideoDetailedCaption
4.39Tau (τ)SAGE
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SAGESetup=Qwen3-VL 235B/A22B-8B2026.01 | 4.39 | 4.33 | 1.32 | |
| SpecVLMTemperature=0, Visual Token Length=25K, Backbone=LLaVA-OneVision-7B2026.02 | 4.15 | — | 1.25 | |
| IVSBTemperature=0, Visual Token Length=25K, Backbone=LLaVA-OneVision-7B2026.02 | 3.88 | — | 1.96 | |
| SD-TreeSetup=Qwen3-VL 235B/A22B-8B2026.01 | 3.58 | 3.99 | 1.22 | |
| SpecVLMSetup=Qwen3-VL 235B/A22B-8B2026.01 | 3.24 | 3.74 | 1.14 | |
| IVSBTemperature=1, Visual Token Length=25K, Backbone=LLaVA-OneVision-7B2026.02 | 2.8 | — | 1.6 | |
| MSDTemperature=0, Visual Token Length=25K, Backbone=LLaVA-OneVision-7B2026.02 | 2.69 | — | 1.47 | |
| MSDTemperature=1, Visual Token Length=25K, Backbone=LLaVA-OneVision-7B2026.02 | 2.2 | — | 1.27 | |
| Vanilla ARSetup=Qwen3-VL 235B/A22B-8B2026.01 | — | 3.27 | — |