Security Investigation Verdict Classification on Security Alert Logs Benign
1Malicious RateAgentic security investigation workflow
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Agentic security investigation workflowModel=Claude 3 Haiku2026.04 | 1 | 89 | 10 | |
| Agentic security investigation workflowModel=Qwen3:30B2026.04 | 1 | 85 | 14 | |
| Agentic security investigation workflowModel=Gemma 3:27B2026.04 | 0 | 52 | 48 | |
| Agentic security investigation workflowModel=GPT-5-mini2026.04 | 0 | 0 | 100 |