Vulnerability Attack Analysis on WebGenBench Internal Tools 1.0
90ASR (CodeQL)Claude-4.5 Opus
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Claude-4.5 OpusEvaluation Protocol=Cross-domain2026.02 | 90 | 93.75 | 90 | 93.33 | |
| GPT-5.2Evaluation Protocol=Cross-domain2026.02 | 88.24 | 87.5 | 83.82 | 75 | |
| GrokEvaluation Protocol=Cross-domain2026.02 | 85 | 78.57 | 81 | 68.18 | |
| ComposerEvaluation Protocol=Cross-domain2026.02 | 83.33 | 87.5 | 71.43 | 87.5 | |
| Gemini-3 ProEvaluation Protocol=Cross-domain2026.02 | 80.77 | 87.5 | 80.77 | 83.33 | |
| Gemini-3 FlashEvaluation Protocol=Cross-domain2026.02 | 80 | 75 | 80 | 75 | |
| GPT-5.2Evaluation Protocol=Held-out (target-domain)2026.02 | 60 | 75 | 60 | 75 | |
| Claude-4.5 OpusEvaluation Protocol=Held-out (target-domain)2026.02 | 50 | 60 | 50 | 60 | |
| ComposerEvaluation Protocol=Held-out (target-domain)2026.02 | 50 | 60 | 50 | 57.14 | |
| Gemini-3 ProEvaluation Protocol=Held-out (target-domain)2026.02 | 40 | 60 | 40 | 44 | |
| GrokEvaluation Protocol=Held-out (target-domain)2026.02 | 33.33 | 50 | 33.33 | 50 | |
| Gemini-3 FlashEvaluation Protocol=Held-out (target-domain)2026.02 | 25 | 60 | 25 | 35 |