Robot Manipulation (Sorting) on AgenticLab Sorting 20 trials across 6 object setups
75Success RateGemini 1.5 Flash
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Gemini 1.5 Flashpipeline=single-VLM2026.02 | 75 | 334 | 67 | |
| Gemini 1.5 Prodetector=LangSAM2026.02 | 70 | 524 | 80.9 | |
| Gemini Roboticspipeline=single-VLM2026.02 | 50 | 382 | 40.6 | |
| Qwen-VL-Pluspipeline=single-VLM2026.02 | 25 | 538 | 43 | |
| Gemini 1.5 Propipeline=single-VLM2026.02 | 10 | 1,283 | 139 | |
| GPT-4odetector=LangSAM2026.02 | 5 | 299 | 49.6 | |
| Qwen-VL-Maxpipeline=single-VLM2026.02 | 0 | — | — | |
| GPT-4o-minidetector=LangSAM2026.02 | 0 | — | — |