Long-horizon embodied manipulation on Long-horizon manipulation tasks (End-effector Variation)
82.5SRRECENT
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| RECENTLLM=Qwen2.5-Coder-7B2026.06 | 82.5 | 89.84 | 2.51 | 0.69 | 2.34 | |
| CAP-CODEXLLM=GPT-5.2-Codex2026.06 | 74.85 | 87.81 | 41.69 | 0.6 | 10.29 | |
| REPAIRAGENTLLM=Qwen2.5-Coder-7B2026.06 | 52.17 | 73.82 | 55.92 | 1.12 | 84.37 | |
| ROBOINSPECTORLLM=Qwen2.5-Coder-7B2026.06 | 39.17 | 64.8 | 52.2 | 5.07 | 31.65 | |
| PROGPROMPTLLM=Qwen2.5-Coder-7B2026.06 | 33.33 | 58.16 | 65.34 | 4.72 | 39.32 | |
| AGENTLESSLLM=Qwen2.5-Coder-7B2026.06 | 26.34 | 49.88 | 74.96 | 5.01 | 42.44 | |
| CAPLLM=Qwen2.5-Coder-7B2026.06 | 21.97 | 36.33 | 79.31 | 3.18 | 55.62 | |
| CAP-CODEV-R1LLM=Qwen2.5-Coder-7B2026.06 | 17.88 | 41.03 | 99.16 | 2.77 | 43.77 | |
| SCOTLLM=Qwen2.5-Coder-7B2026.06 | 15.83 | 54.19 | 85.35 | 1.8 | 41.87 |