Video Understanding on SEED-Bench Video Understanding
74.12AccuracyBaseline
Evaluation Results
| Method | Links | |
|---|---|---|
| BaselineBackbone=Qwen2.5-VL-7B-Instruct, Token Retention=25%2026.02 | 74.12 | |
| IDPrunerBackbone=Qwen2.5-VL-7B-Instruct, Token Retention=25%2026.02 | 72.68 | |
| SCOPEBackbone=Qwen2.5-VL-7B-Instruct, Token Retention=25%2026.02 | 72.63 | |
| DivPruneBackbone=Qwen2.5-VL-7B-Instruct, Token Retention=25%2026.02 | 72.11 | |
| VisionZipBackbone=Qwen2.5-VL-7B-Instruct, Token Retention=25%2026.02 | 72.11 | |
| HiPruneBackbone=Qwen2.5-VL-7B-Instruct, Token Retention=25%2026.02 | 71.97 | |
| VisPrunerBackbone=Qwen2.5-VL-7B-Instruct, Token Retention=25%2026.02 | 71.89 | |
| InternVL-2.5-26B (Reproduction)Backbone=InternVL-2.5-26B, Red. Ratio=0.00%, Latency=2.9142, Speedup=-2025.12 | 71.41 | |
| InternVL-2.5-26B + LUVCBackbone=InternVL-2.5-26B, Red. Ratio=61.32%, Latency=1.2255, Speedup=2.38x2025.12 | 71.28 | |
| DARTBackbone=Qwen2.5-VL-7B-Instruct, Token Retention=25%2026.02 | 71.27 | |
| InternVL-2.5-26B + VTWBackbone=InternVL-2.5-26B, Red. Ratio=66.67%, Latency=1.2586, Speedup=2.32x2025.12 | 70.96 | |
| VisionSelectorBackbone=Qwen2.5-VL-7B-Instruct, Token Retention=25%2026.02 | 70.75 | |
| InternVL-2.5-26B + PACTBackbone=InternVL-2.5-26B, Red. Ratio=59.61%, Latency=1.3548, Speedup=2.15x2025.12 | 70.24 | |
| FastVBackbone=Qwen2.5-VL-7B-Instruct, Token Retention=25%2026.02 | 69.22 | |
| InternVL-2.5-8B (Reproduction)Backbone=InternVL-2.5-8B, Red. Ratio=0.00%, Latency=1.2480, Speedup=-2025.12 | 66.41 | |
| InternVL-2.5-8B + LUVCBackbone=InternVL-2.5-8B, Red. Ratio=62.12%, Latency=0.6293, Speedup=1.98x2025.12 | 65.85 | |
| InternVL-2.5-8B + VTWBackbone=InternVL-2.5-8B, Red. Ratio=56.25%, Latency=0.6482, Speedup=1.93x2025.12 | 65.45 | |
| InternVL-2.5-8B + PACTBackbone=InternVL-2.5-8B, Red. Ratio=55.41%, Latency=0.6949, Speedup=1.80x2025.12 | 65.18 | |
| Dream-VLLLM Backbone=Dream 7B, Open Data=true2025.12 | 58.1 | |
| LlavaOV-7B + LUVCBackbone=LlavaOV-7B, Red. Ratio=62.32%, Latency=0.2583, Speedup=2.01x2025.12 | 57.36 | |
| MAmmoTH-VLLLM Backbone=Qwen2.5-7B, Open Data=true2025.12 | 57.1 | |
| LLaVA-OVLLM Backbone=Qwen2-7B, Open Data=true2025.12 | 56.9 | |
| LlavaOV-7BBackbone=LlavaOV-7B, Red. Ratio=0.00%, Latency=0.5194, Speedup=-2025.12 | 56.9 | |
| LlavaOV-7B + PACTBackbone=LlavaOV-7B, Red. Ratio=55.88%, Latency=0.2711, Speedup=1.92x2025.12 | 56.8 | |
| LlavaOV-7B + FastVBackbone=LlavaOV-7B, Red. Ratio=42.86%, Latency=0.2693, Speedup=1.93x2025.12 | 56.72 | |
| LLaDA-VLLM Backbone=LLaDA 8B, Open Data=true2025.12 | 53.7 | |
| LlavaOV-7B + VTWBackbone=LlavaOV-7B, Red. Ratio=57.14%, Latency=0.2524, Speedup=2.06x2025.12 | 48.44 | |
| LlavaOV-0.5B + LUVCBackbone=LlavaOV-0.5B, Red. Ratio=58.46%, Latency=0.0610, Speedup=1.27x2025.12 | 44.96 | |
| LlavaOV-0.5B + PACTBackbone=LlavaOV-0.5B, Red. Ratio=48.05%, Latency=0.0642, Speedup=1.20x2025.12 | 44.8 | |
| LlavaOV-0.5BBackbone=LlavaOV-0.5B, Red. Ratio=0.00%, Latency=0.0772, Speedup=-2025.12 | 44.2 | |
| LlavaOV-0.5B (Reproduction)Backbone=LlavaOV-0.5B, Red. Ratio=0.00%, Latency=0.0772, Speedup=-2025.12 | 44.16 | |
| LlavaOV-0.5B + VTWBackbone=LlavaOV-0.5B, Red. Ratio=54.17%, Latency=0.0596, Speedup=1.30x2025.12 | 43.86 | |
| LlavaOV-0.5B + FastVBackbone=LlavaOV-0.5B, Red. Ratio=41.67%, Latency=0.0711, Speedup=1.00x2025.12 | 43.55 |