Vulnerability Attack Analysis on WebGenBench Blogging target-domain and cross-domain 1.0
100ASR (CodeQL)Claude-4.5 Opus
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Claude-4.5 OpusEvaluation Protocol=Held-out (target-domain)2026.02 | 100 | 100 | 100 | 100 | |
| Gemini-3 FlashEvaluation Protocol=Held-out (target-domain)2026.02 | 100 | 100 | 100 | 100 | |
| GPT-5.2Evaluation Protocol=Cross-domain2026.02 | 80.65 | 87.5 | 77.94 | 77.78 | |
| GPT-5.2Evaluation Protocol=Held-out (target-domain)2026.02 | 80 | 80 | 80 | 65 | |
| Gemini-3 ProEvaluation Protocol=Held-out (target-domain)2026.02 | 80 | 80 | 80 | 80 | |
| Claude-4.5 OpusEvaluation Protocol=Cross-domain2026.02 | 76.92 | 84.62 | 76.92 | 84 | |
| ComposerEvaluation Protocol=Cross-domain2026.02 | 76.92 | 81.82 | 69.23 | 81.25 | |
| GrokEvaluation Protocol=Held-out (target-domain)2026.02 | 75 | 60 | 75 | 57.14 | |
| GrokEvaluation Protocol=Cross-domain2026.02 | 75 | 77.78 | 71.43 | 66.67 | |
| Gemini-3 ProEvaluation Protocol=Cross-domain2026.02 | 69.23 | 81.25 | 69.23 | 72 | |
| ComposerEvaluation Protocol=Held-out (target-domain)2026.02 | 66.67 | 75 | 50 | 75 | |
| Gemini-3 FlashEvaluation Protocol=Cross-domain2026.02 | 42.86 | 63.64 | 42.86 | 52 |