World-state tracking on ToolDAGWorld Procedural
49.1World-State Accuracy (AH)(c+d)-only
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| (c+d)-onlyStructural variants=criticality + dependency only2026.06 | 49.1 | 15.9 | |
| -u (no uncertainty)Structural variants=no uncertainty2026.06 | 48.8 | 8.2 | |
| EnvProbe-SimpleProbe strategy=Simple2026.06 | 43.1 | 20.5 | |
| Periodic-ProbeProbe strategy=Periodic2026.06 | 31.3 | — |