Long-horizon embodied manipulation on Long-horizon manipulation tasks Kinematic Variation
73Success Rate (SR)RECENT
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| RECENTLLM=Qwen2.5-Coder-7B2026.06 | 73 | 81.76 | 7.85 | 0.72 | 1.77 | |
| CAP-CODEXLLM=GPT-5.2-Codex2026.06 | 67.5 | 80.45 | 84.29 | 1.23 | 7.99 | |
| REPAIRAGENTLLM=Qwen2.5-Coder-7B2026.06 | 42 | 64.05 | 54.17 | 1.73 | 50.53 | |
| ROBOINSPECTORLLM=Qwen2.5-Coder-7B2026.06 | 41 | 60.83 | 179.4 | 1.15 | 59.98 | |
| PROGPROMPTLLM=Qwen2.5-Coder-7B2026.06 | 31.5 | 53.89 | 86.47 | 3.28 | 30.74 | |
| AGENTLESSLLM=Qwen2.5-Coder-7B2026.06 | 22.5 | 42.7 | 128.08 | 2.19 | 58.27 | |
| CAP-CODEV-R1LLM=Qwen2.5-Coder-7B2026.06 | 20 | 36.1 | 109.38 | 4.13 | 17.5 | |
| CAPLLM=Qwen2.5-Coder-7B2026.06 | 17 | 41.11 | 102.01 | 4.72 | 25.65 | |
| SCOTLLM=Qwen2.5-Coder-7B2026.06 | 14.5 | 38.38 | 107.08 | 4.4 | 28.86 |