Household Planning on Household Domains
96.7Success Rate (PC Assembly)DUPLEX Full System
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| DUPLEX Full SystemMethod Architecture=Full System, Backbone Model=Qwen3-8B + GPT-4o2026.03 | 96.7 | — | 98 | 70.7 | 68.7 | 83.5 | |
| DUPLEX Fast SystemMethod Architecture=Fast System, Backbone Model=Qwen3-8B2026.03 | 76.7 | — | 70 | 30.7 | 26 | 50.9 | |
| LLM+PBackbone Model=GPT-4o2026.03 | 76 | — | 4 | 0 | 0 | 20 | |
| LLM-as-PlannerBackbone Model=GPT-4o2026.03 | 70 | — | 38.7 | 0 | 0 | 27.2 |