SQL Agent data leakage evaluation on Employee Medium
100Benign Accuracy (BA)gpt-4.1-mini
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| gpt-4.1-miniSetup=Orchestrator with SQL Agent, Attack Type=Explicit Attacks2026.02 | 100 | 73.6 | 4 | |
| o4-miniSetup=Orchestrator with SQL Agent, Attack Type=Explicit Attacks2026.02 | 100 | 84.6 | — | |
| claude-sonnet-4Setup=Orchestrator with SQL Agent, Attack Type=Explicit Attacks2026.02 | 100 | 93.6 | — | |
| gemini-2.5-flashSetup=Orchestrator with SQL Agent, Attack Type=Explicit Attacks2026.02 | 100 | 61.8 | 17 | |
| gpt-4.1-mini2026.02 | 100 | 76.2 | 7 | |
| o4-mini2026.02 | 100 | 84 | — | |
| claude-sonnet-42026.02 | 100 | 95.2 | — | |
| gemini-2.5-flash2026.02 | 100 | 63.6 | 20 | |
| gpt-4.1Setup=Orchestrator with SQL Agent, Attack Type=Explicit Attacks2026.02 | 98 | 63.4 | 17 | |
| gpt-4.12026.02 | 98 | 64.6 | 34 |