Pick on Real Robot Manipulation real-world
86Success RateClaude Sonnet 4.5
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Claude Sonnet 4.5zero-shot=true, Toolshed enhancement=true2025.12 | 86 | 30 | |
| SpaceToolsinteractive RL=true2025.12 | 86 | 10 | |
| GPT-5zero-shot=true, Toolshed enhancement=true2025.12 | 71 | 36 | |
| π0.5type=Vision-Language Action Model2025.12 | 0 | 1 | |
| Qwen2.5-VL-3Bzero-shot=true, Toolshed enhancement=true2025.12 | 0 | — |