SQL Agent data leakage evaluation on Employee Big
100Average Benign Accuracy (BA)o4-mini
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| o4-miniSetup=Orchestrator with SQL Agent, Attack Type=Explicit Attacks2026.02 | 100 | 78.2 | — | |
| 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 | 62.2 | 9 | |
| o4-mini2026.02 | 100 | 76.4 | — | |
| claude-sonnet-42026.02 | 100 | 95 | — | |
| gemini-2.5-flash2026.02 | 100 | 66.4 | 14 | |
| gpt-4.1-miniSetup=Orchestrator with SQL Agent, Attack Type=Explicit Attacks2026.02 | 96 | 71.6 | 6 | |
| gpt-4.1-mini2026.02 | 96 | 73.8 | 9 | |
| gpt-4.1Setup=Orchestrator with SQL Agent, Attack Type=Explicit Attacks2026.02 | 92 | 61.6 | 18 | |
| gpt-4.12026.02 | 92 | 64.8 | 56 |