Safety evaluation against adversarial collusion on 50 clinical prescribing cases (test)
0ASRVerifier Agent
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Verifier AgentMode=With Verifier2025.12 | 0 | 0 | 100 | 100 | |
| Unscripted (C)Mode=Unscripted (Consensus-biased)2025.12 | 98 | 98 | 2 | 0 | |
| Unscripted (N)Mode=Unscripted (Neutral)2025.12 | 98 | 98 | 2 | 0 | |
| Scripted ModeMode=Scripted2025.12 | 100 | 100 | 0 | 0 |