Benchmarks
347 sources · 1,258 models × 3,038 benchmarks · 158,912 scoresCyberGym-E2E evaluates the complete vulnerability lifecycle: autonomous discovery, proof-of-concept generation, patch generation, regression testing, and target-vulnerability repair. This overview definition has no direct score rows; the official 615-task and 920-task leaderboard measurements use separate factor families.
No signal or difficulty estimate. Held out of the fit by design.
This is a benchmark-family page. Scores are recorded on the component benchmarks below rather than directly on the parent, because the components use different score scales and directions.
| Metric | Models | Leader | Direction | Date |
|---|---|---|---|---|
| CyberGym-E2E-AA (Artificial Analysis) | 22 | Kimi K3.1 Pro94.7 | higher is better | 2026-10-06 |