Object Recognition on EgoPro-Bench Object
94PrecisionProAct-Stream
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| ProAct-StreamTraining Strategy=RL, Think Data usage=W Think2026.05 | 94 | 90.22 | 91.19 | 89.89 | 87.01 | |
| Qwen3VL-30B-A3B2026.05 | 90.97 | 54.75 | 62.28 | 82.51 | 53.24 | |
| ProAct-StreamTraining Strategy=SFT2026.05 | 90.78 | 87.24 | 88 | 88.07 | 84.01 | |
| ProAct-StreamTraining Strategy=RL, Think Data usage=W/O Think2026.05 | 90.78 | 87.36 | 88.07 | 87.94 | 84.12 | |
| Qwen2.5VL-72B2026.05 | 90.67 | 83.17 | 84.01 | 83.03 | 77.37 | |
| Qwen3VL-4B2026.05 | 79.11 | 77.61 | 75.01 | 79.54 | 67.7 | |
| Qwen2.5VL-32B2026.05 | 68.93 | 48.52 | 52.97 | 73.35 | 46.5 | |
| Qwen3VL-8b2026.05 | 59.45 | 47.96 | 49.35 | 71.85 | 44.57 | |
| VideoChat-R1.52026.05 | 56.8 | 42.06 | 40.93 | 65.2 | 33.87 | |
| VideoRFT-7B2026.05 | 43.64 | 61.75 | 44.15 | 55.4 | 35.52 | |
| Qwen2.5VL-7B2026.05 | 40.72 | 44.25 | 36.45 | 58.91 | 30.29 | |
| TimeChatOnline-7B2026.05 | 18.83 | 22.83 | 18.56 | 55.16 | 15.48 | |
| Qwen2.5vl-3B2026.05 | 5.33 | 4.02 | 3.67 | 52.49 | 2.69 |