Fix instruction duration unit in estimated success probability calculation

CHANGELOG.md

@@ -11,6 +11,7 @@ This project adheres to [Semantic Versioning], with the exception that minor rel
 
 ### Changed
 
+- 🐛 Fix instruction duration unit in estimated success probability calculation ([#445])([**@Shaobo-Zhou])
 - ✨ Remove support for custom names of trained models ([#489]) ([**@bachase**])
 - 🔥 Drop support for x86 macOS systems ([#421]) ([**@denialhaag**])
 

src/mqt/predictor/reward.py

@@ -203,8 +203,10 @@ def estimated_success_probability(qc: QuantumCircuit, device: Target, precision:
                 if first_qubit_idx not in active_qubits:
                     continue
 
+                dt = device.dt  # instruction durations are stored in unit dt
                 res *= np.exp(
                     -instruction.duration
+                    * dt
                     / min(device.qubit_properties[first_qubit_idx].t1, device.qubit_properties[first_qubit_idx].t2)
                 )
                 continue