Planning on IPC domains (test)
100BarmanLLM+P
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| LLM+PBackbone=GPT-4o2026.03 | 100 | 90 | 15 | 100 | 85 | 90 | 90 | 5 | 71.9 | |
| DUPLEXSystem Variant=Fast System, Backbone=Qwen3-8B2026.03 | 100 | 100 | 30 | 100 | 100 | 100 | 100 | 40 | 83.8 | |
| DUPLEXSystem Variant=Full System, Backbone=Qwen3-8B + GPT-4o2026.03 | 100 | 100 | 85 | 100 | 100 | 100 | 100 | 95 | 97.5 | |
| LLM-as-PlannerBackbone=GPT-4o2026.03 | 0 | 30 | 0 | 50 | 0 | 0 | 15 | 0 | 11.9 |