LLM Recovery on Navigation (Official)
100Success RateRETRY-ONLY
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| RETRY-ONLYRegime=Official2026.05 | 100 | 10 | 6 | 0 | 6,920.2 | 0 | — | |
| Coarse-State-RetryRegime=Official2026.05 | 100 | 5 | 0 | 100 | 4,735.49 | 100 | — | |
| Comp-EntryOnlyRegime=Official2026.05 | 100 | 5 | 0 | 100 | 4,640.04 | 100 | — | |
| COMP-FROZENRegime=Official2026.05 | 100 | 5 | 0 | 100 | 4,841.55 | 100 | — | |
| LangGraph-SemiRealRuntime family=Semi-real LangGraph-based, Primary use=strong external baseline2026.05 | 100 | 1 | — | — | 4.33 | — | — | |
| DARTRuntime family=Proposed runtime, Primary use=recoverability controller2026.05 | 100 | 5.5 | — | — | 9,619.25 | — | — |