Offline Trajectory Selection on Real-world sidewalk dataset (hard)
1.16ADEGemini 3 Flash
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Gemini 3 FlashTop-K=182026.06 | 1.16 | 0.39 | 0.66 | 16.76 | 6.16 | 15.7 | 2,744 | 62 | 5,957 | 8.1 | |
| Gemini 2.5 FlashTop-K=182026.06 | 1.2 | 0.41 | 0.69 | 17.04 | 6.32 | 16.14 | 1,902 | 82 | 4,134 | 13.2 | |
| Qwen2.5-VL-72BTop-K=182026.06 | 1.2 | 0.4 | 0.68 | 17.11 | 6.01 | 16.18 | 2,250 | 43 | 2,293 | 11.8 | |
| Gemini 2.5 Flash LiteTop-K=182026.06 | 1.21 | 0.4 | 0.68 | 17.43 | 6.2 | 16.5 | 1,902 | 85 | 1,994 | 1.7 | |
| Gemini Robotics ER 1.5Top-K=182026.06 | 1.21 | 0.41 | 0.69 | 16.78 | 6.06 | 15.86 | 1,902 | 73 | 3,372 | 14.4 | |
| GPT-5 MiniTop-K=182026.06 | 1.47 | 0.48 | 0.82 | 17.53 | 5.05 | 18 | 2,181 | — | 3,196 | 16.6 | |
| Planner ArgmaxTop-K=182026.06 | 1.64 | 0.64 | 1.06 | 16.08 | 4.88 | 16.38 | — | — | — | — | |
| GPT-5 NanoTop-K=182026.06 | 1.77 | 0.54 | 0.98 | 17.53 | 4.43 | 17.33 | 2,334 | — | 5,053 | 22 | |
| Qwen2.5-VL-32BTop-K=182026.06 | 1.81 | 0.55 | 0.99 | 17.2 | 4.12 | 17.41 | 2,250 | 52 | 2,302 | 13.1 | |
| Qwen2.5-VL-7BTop-K=182026.06 | 2.27 | 0.69 | 1.25 | 16.46 | 3.33 | 19.45 | 2,250 | 48 | 2,298 | 2.7 |