Tone-controlled captioning on RoadTones Eval 1.0 (test)
72.7P ScoreROADTONES-VL w-CoT
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| ROADTONES-VL w-CoTParams=8B, base_model=Qwen3-VL ft.2026.05 | 72.7 | 81.5 | 77.1 | 98.2 | 74.5 | 94.5 | 89.1 | 83.1 | 57.7 | 70.4 | |
| ROADTONES-VL w/o-CoTParams=8B, base_model=Qwen3-VL ft.2026.05 | 72.7 | 81.6 | 77.2 | 98.1 | 74.2 | 94.1 | 88.8 | 83 | 57.2 | 70.1 | |
| Single-RunParams=8B, type=ablation2026.05 | 72.1 | 80.8 | 76.5 | 97.9 | 73.2 | 94.2 | 88.4 | 82.4 | 56.3 | 69.4 | |
| Single-CapParams=8B, type=ablation2026.05 | 70.9 | 79.5 | 75.2 | 98.4 | 69.5 | 92.1 | 86.6 | 80.9 | 53 | 67 | |
| Gemini-2.5-promode=closed-source general-purpose2026.05 | 68.5 | 72.5 | 70.5 | 95.8 | 74.9 | 91.7 | 87.5 | 79 | 59.4 | 69.2 | |
| RoadTones-CapParams=8B, type=ablation2026.05 | 65.4 | 73.1 | 69.2 | 98.1 | 73.3 | 80.6 | 84 | 76.6 | 54.2 | 65.4 | |
| GPT-5mode=closed-source general-purpose2026.05 | 63.1 | 65 | 64.1 | 95.7 | 74.3 | 87.3 | 85.8 | 75 | 55.7 | 65.4 | |
| Qwen3-VLParams=8B, mode=open-source general-purpose2026.05 | 60.7 | 57.9 | 59.3 | 82.6 | 69 | 17.5 | 56.4 | 57.9 | 52.8 | 55.4 | |
| RoadSocial-CapParams=8B, type=ablation2026.05 | 60 | 64.9 | 62.4 | 93.9 | 68.1 | 77.7 | 79.9 | 71.2 | 46.2 | 58.7 | |
| MiniCPM-V 4.5Params=8B, mode=open-source general-purpose2026.05 | 59.9 | 55 | 57.4 | 79.3 | 69.6 | 18.3 | 55.7 | 56.6 | 47.6 | 52.1 | |
| InternVL-3.5Params=8B, mode=open-source general-purpose2026.05 | 50.3 | 52.4 | 51.3 | 77.8 | 64.8 | 22.9 | 55.2 | 53.3 | 44.1 | 48.7 | |
| RoadSocialParams=7B, mode=open-source driving-specific2026.05 | 40.8 | 33.7 | 37.2 | 63.4 | 73 | 52.2 | 62.9 | 50.1 | 44.5 | 47.3 | |
| VideoLLaMA3Params=7B, mode=open-source general-purpose2026.05 | 40.8 | 43.6 | 42.2 | 68.9 | 64 | 12 | 48.3 | 45.3 | 41.7 | 43.5 | |
| Qwen2.5-VLParams=7B, mode=open-source general-purpose2026.05 | 37.5 | 37.1 | 37.3 | 65 | 66 | 18.9 | 50 | 43.7 | 48.3 | 46 | |
| RoboTron-DriveParams=8B, mode=open-source driving-specific2026.05 | 27.8 | 32.9 | 30.3 | 61.1 | 58.1 | 32.9 | 47.8 | 39.1 | 41.5 | 40.3 | |
| DolphinsParams=9B, mode=open-source driving-specific2026.05 | 19.5 | 32.6 | 26.1 | 63 | 56.5 | 37.3 | 52.3 | 39.2 | 40.2 | 39.7 |