CrypFormBench - Correction: leaderboard
Formal-model correction on CrypFormBench (2026), a formal cryptographic analysis benchmark over seven verifier languages: the model repairs formal models with injected errors (weight 0.4) or false security claims (weight 0.6), scored 0 to 100 and averaged over languages. Higher is better.
Metric: Correction score (0-100) on CrypFormBench (700 instances over 677 cryptographic schemes and 7 formal verifier languages: Scyther, Tamarin, AVISPA, ProVerif, Maude-NPA, CryptoVerif, EasyCrypt): repair a formal model with injected errors or false security claims, combining error and false-claim repair scores 0.4 and 0.6, averaged over languages; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek R1 | 56.9 | #473 |
| 2 | GPT-4o | 54.7 | #741 |
| 3 | GPT-4o Mini | 44.8 | #1623 |
| 4 | Gemini 2.5 Pro | 35.6 | #266 |
| 5 | GLM-4 | 14.5 | #1666 |
| 6 | Grok 3 | 0 | #602 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=crypformbench-correction · How It Works · Data refreshed daily, snapshot 2026-11-13.