Method | Normal �Success | Expected Match | Unsafe Pass-�through | Runtime �Failure | Safety �Events | Speed �Clamp | Avg. Runtime (s) |
Direct Execution �Baseline | 100.0% | 33.3% | 55.0% (55/100) | 30.0% (45) | 0 | 0 | 7.8283 |
Schema-only �Baseline | 100.0% | 50.0% | 55.0% (55/100) | 13.3% (20) | 0 | 0 | 7.7992 |
DSL-Guided �Proposed | 100.0% | 100.0% | 0.0% (0/100) | 0.0% (0) | 30 | 30 | 6.5004 |
Table 1. Comparative evaluation of AMR DSL execution methods on the 150-scenario benchmark.
For reproducibility, we define three command handling prompts corresponding to the three evaluated runtime strategies: direct execution, schema only validation, and DSL guided runtime validation. The direct prompt attempts to execute the command without symbolic mediation. The schema only prompt checks structural validity but does not enforce semantic or safety constraints. The DSL guided prompt converts the command into a symbolic DSL plan and applies schema validation, semantic validation, and safety guard enforcement before execution. [Prompts]