Failure prediction stopping on PCS
87.4Max Accuracyα-STOP
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| α-STOPPredictor=attention-based failure predictor2026.06 | 87.4 | 0.734 | 0.0034 | |
| LLM Judge (best)Predictor=attention-based failure predictor2026.06 | 86.2 | — | — | |
| Plug-in ThresholdPredictor=attention-based failure predictor2026.06 | 81.3 | 0.694 | 0.0251 | |
| ALERT*Predictor=attention-based failure predictor2026.06 | 80.8 | 0.681 | 0.0317 | |
| FIRMBOUNDPredictor=attention-based failure predictor2026.06 | 73.5 | 0.669 | 0.053 | |
| EndtoEndRLPredictor=attention-based failure predictor2026.06 | 59.7 | 0.568 | 0.1353 |