Streaming Video Understanding on RVS-Ego
63.7AccuracyReKV
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ReKVBase Model=LLaVA-OV-7B, Upper Bound=True2026.01 | 63.7 | 4 | |
| ReKVCPU offloading=true2025.10 | 63.7 | 4 | |
| HERMESBase Model=LLaVA-OV-7B, Token Budget=6K tokens2026.01 | 60.3 | 4 | |
| StreamingTOMBase Model=LLaVA-OV-7B2026.01 | 58.3 | 3.9 | |
| HERMESBase Model=LLaVA-OV-7B, Token Budget=4K tokens2026.01 | 58.3 | 3.9 | |
| StreamingTOM2025.10 | 58.3 | 3.9 | |
| InfiniPot-VBase Model=LLaVA-OV-7B2026.01 | 57.9 | 3.5 | |
| InfiniPot-V2025.10 | 57.9 | 3.5 | |
| StreamMemBase Model=LLaVA-OV-7B2026.01 | 57.6 | 3.8 | |
| StreamMem2025.10 | 57.6 | 3.8 | |
| Flash-VStreamBase Model=LLaVA-OV-7B2026.01 | 57 | 4 | |
| Flash-VStream2025.10 | 57 | 4 | |
| LLaVA-OV-7BBase Model=LLaVA-OV-7B2026.01 | 56.2 | 3.7 | |
| ReKV w/o off.Base Model=LLaVA-OV-7B2026.01 | 55.8 | 3.3 | |
| ReKVCPU offloading=false2025.10 | 55.8 | 3.3 | |
| ReKVBase Model=LLaVA-OV-0.5B, Upper Bound=True2026.01 | 54.7 | 3.9 | |
| HERMESBase Model=LLaVA-OV-0.5B, Token Budget=6K tokens2026.01 | 53 | 3.8 | |
| HERMESBase Model=LLaVA-OV-0.5B, Token Budget=4K tokens2026.01 | 52.7 | 3.8 | |
| LLaVA-OV-0.5BBase Model=LLaVA-OV-0.5B2026.01 | 51.8 | 3.7 | |
| ViCoStreamBackbone=Qwen2.5-VL-7B, acn=4, ur=32, s=0.25 (↓48.0%)2026.06 | 36.71 | — | |
| ViCoStreamBackbone=Qwen2.5-VL-7B, acn=4, ur=64, s=0.25 (↓48.0%)2026.06 | 36.68 | — | |
| ViCoStreamBackbone=Qwen2.5-VL-7B, acn=8, ur=64, s=0.1 (↓70.5%)2026.06 | 35.63 | — | |
| ViCoStreamBackbone=Qwen2.5-VL-3B, acn=8, ur=64, s=0.1 (↓73.0%)2026.06 | 35.43 | — | |
| ViCoStreamBackbone=Qwen2.5-VL-3B, acn=4, ur=64, s=0.25 (↓57.6%)2026.06 | 35.22 | — | |
| ViCoStreamBackbone=Qwen2.5-VL-7B, acn=8, ur=32, s=0.1 (↓70.5%)2026.06 | 34.88 | — | |
| Qwen2.5-VL-7BMode=Traditional Streaming Prefill (Baseline), acn=∞, ur=∞, s=–2026.06 | 34.86 | — | |
| Qwen2.5-VL-3BMode=Traditional Streaming Prefill (Baseline), acn=∞, ur=∞, s=–2026.06 | 34.59 | — | |
| ViCoStreamBackbone=Qwen2.5-VL-3B, acn=4, ur=64, s=0.1 (↓72.9%)2026.06 | 34.4 | — | |
| ViCoStreamBackbone=Qwen2.5-VL-7B, acn=8, ur=32, s=0.25 (↓48.4%)2026.06 | 34.4 | — | |
| ViCoStreamBackbone=Qwen2.5-VL-7B, acn=8, ur=64, s=0.25 (↓48.4%)2026.06 | 34.31 | — | |
| ViCoStreamBackbone=Qwen2.5-VL-3B, acn=8, ur=64, s=0.25 (↓57.3%)2026.06 | 34.12 | — | |
| ViCoStreamBackbone=Qwen2.5-VL-3B, acn=8, ur=32, s=0.1 (↓72.9%)2026.06 | 32.7 | — | |
| ViCoStreamBackbone=Qwen2.5-VL-7B, acn=4, ur=64, s=0.1 (↓70.6%)2026.06 | 32.7 | — | |
| ViCoStreamBackbone=Qwen2.5-VL-3B, acn=4, ur=32, s=0.25 (↓57.6%)2026.06 | 32.29 | — | |
| ViCoStreamBackbone=Qwen2.5-VL-3B, acn=8, ur=32, s=0.25 (↓57.3%)2026.06 | 32.18 | — | |
| ViCoStreamBackbone=Qwen2.5-VL-7B, acn=4, ur=32, s=0.1 (↓70.6%)2026.06 | 32.12 | — | |
| ViCoStreamBackbone=Qwen2.5-VL-3B, acn=4, ur=32, s=0.1 (↓72.9%)2026.06 | 30.99 | — |