Vulnerability Fixing on curated CVE dataset
44Fix AccuracyOpenAI GPT-5.5.
Evaluation results
| Method | Links | |
|---|---|---|
| OpenAI GPT-5.5.Setting=Baseline2026.06 | 44 | |
| role-based agentic workflow (Workflow 1)CodeQL=without CodeQL, Planner=included2026.06 | 19 | |
| role-based agentic workflow (Workflow 2)CodeQL=with CodeQL, Planner=included2026.06 | 13 | |
| role-based agentic workflow (Workflow 1)CodeQL=without CodeQL, Planner=without Planner2026.06 | 13 | |
| role-based agentic workflow (Workflow 2)CodeQL=with CodeQL, Planner=without Planner2026.06 | 6 |