Hazard Detection and Temporal Localization on HomeSafe-Bench 1.0 (test)
97.03HDRInternVL3.5-8B
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| InternVL3.5-8BModel Access Type=Open-Source, Evaluation Protocol=Zero-shot, Input Resolution=448x448, Sampling Rate=10 FPS, Sliding Window=2s length, 1.5s stride2026.03 | 97.03 | 38.35 | 53.88 | 22.15 | 9.82 | 5.48 | 8.68 | 28.42 | |
| Claude-Opus-4.1Model Access Type=Closed-Source, Evaluation Protocol=Zero-shot, Input Resolution=448x448, Sampling Rate=10 FPS, Sliding Window=2s length, 1.5s stride2026.03 | 93.61 | 25.12 | 66.89 | 14.84 | 5.48 | 3.2 | 9.59 | 18.38 | |
| InternVL3.5-4BModel Access Type=Open-Source, Evaluation Protocol=Zero-shot, Input Resolution=448x448, Sampling Rate=10 FPS, Sliding Window=2s length, 1.5s stride2026.03 | 91.78 | 35.07 | 53.65 | 13.01 | 10.73 | 8.45 | 14.16 | 20.49 | |
| MiniCPM-V-4.5Model Access Type=Open-Source, Evaluation Protocol=Zero-shot, Input Resolution=448x448, Sampling Rate=10 FPS, Sliding Window=2s length, 1.5s stride2026.03 | 89.73 | 30.03 | 57.53 | 16.21 | 5.48 | 5.25 | 15.53 | 20.26 | |
| Qwen3-VL-8BModel Access Type=Open-Source, Evaluation Protocol=Zero-shot, Input Resolution=448x448, Sampling Rate=10 FPS, Sliding Window=2s length, 1.5s stride2026.03 | 89.04 | 40 | 47.03 | 18.04 | 9.36 | 8.22 | 17.35 | 24.77 | |
| VITA-1.5Model Access Type=Open-Source, Evaluation Protocol=Zero-shot, Input Resolution=448x448, Sampling Rate=10 FPS, Sliding Window=2s length, 1.5s stride2026.03 | 88.81 | 10.54 | 76.71 | 7.53 | 1.14 | 0.68 | 13.93 | 8.28 | |
| InternVL3.5-1BModel Access Type=Open-Source, Evaluation Protocol=Zero-shot, Input Resolution=448x448, Sampling Rate=10 FPS, Sliding Window=2s length, 1.5s stride2026.03 | 87.9 | 36.36 | 43.61 | 9.59 | 8.22 | 14.16 | 24.43 | 17.24 | |
| HD-GuardModel Access Type=Open-Source, Evaluation Protocol=Zero-shot, Input Resolution=448x448, Sampling Rate=10 FPS, Sliding Window=2s length, 1.5s stride2026.03 | 86.53 | 49.34 | 24.89 | 15.07 | 11.87 | 15.75 | 32.19 | 24.94 | |
| MiniCPM-o-2.6Model Access Type=Open-Source, Evaluation Protocol=Zero-shot, Input Resolution=448x448, Sampling Rate=10 FPS, Sliding Window=2s length, 1.5s stride2026.03 | 85.16 | 28.69 | 56.62 | 9.36 | 7.31 | 7.76 | 18.95 | 14.95 | |
| InternVL3.5-2BModel Access Type=Open-Source, Evaluation Protocol=Zero-shot, Input Resolution=448x448, Sampling Rate=10 FPS, Sliding Window=2s length, 1.5s stride2026.03 | 78.08 | 47.95 | 29.22 | 15.75 | 10.96 | 10.73 | 33.33 | 23.92 | |
| GPT-5.1Model Access Type=Closed-Source, Evaluation Protocol=Zero-shot, Input Resolution=448x448, Sampling Rate=10 FPS, Sliding Window=2s length, 1.5s stride2026.03 | 75.11 | 43.77 | 37.9 | 13.93 | 9.82 | 9.13 | 29.22 | 21.12 | |
| MiniCPM-o-4.5Model Access Type=Open-Source, Evaluation Protocol=Zero-shot, Input Resolution=448x448, Sampling Rate=10 FPS, Sliding Window=2s length, 1.5s stride2026.03 | 74.43 | 34.05 | 44.98 | 12.79 | 7.31 | 5.25 | 29.68 | 17.75 | |
| LLaVA-OV-7BModel Access Type=Open-Source, Evaluation Protocol=Zero-shot, Input Resolution=448x448, Sampling Rate=10 FPS, Sliding Window=2s length, 1.5s stride2026.03 | 70.32 | 40.26 | 31.51 | 10.96 | 8.22 | 9.13 | 40.18 | 17.35 | |
| Qwen3-VL-2BModel Access Type=Open-Source, Evaluation Protocol=Zero-shot, Input Resolution=448x448, Sampling Rate=10 FPS, Sliding Window=2s length, 1.5s stride2026.03 | 64.16 | 48.4 | 17.12 | 8.68 | 6.16 | 16.44 | 51.6 | 15.87 | |
| Qwen3-VL-4BModel Access Type=Open-Source, Evaluation Protocol=Zero-shot, Input Resolution=448x448, Sampling Rate=10 FPS, Sliding Window=2s length, 1.5s stride2026.03 | 51.37 | 65.78 | 10.73 | 7.31 | 9.36 | 17.12 | 55.48 | 16.27 | |
| VideoLlama3-7BModel Access Type=Open-Source, Evaluation Protocol=Zero-shot, Input Resolution=448x448, Sampling Rate=10 FPS, Sliding Window=2s length, 1.5s stride2026.03 | 45.99 | 59.09 | 7.32 | 3.83 | 7.32 | 16.38 | 65.16 | 11.59 | |
| LLaVA-OV-0.5BModel Access Type=Open-Source, Evaluation Protocol=Zero-shot, Input Resolution=448x448, Sampling Rate=10 FPS, Sliding Window=2s length, 1.5s stride2026.03 | 12.93 | 52.63 | 5.1 | 3.74 | 3.06 | 0.34 | 87.76 | 5.36 |