CrypFormBench - Interpretation: leaderboard
Formal-model interpretation on CrypFormBench (2026), a formal cryptographic analysis benchmark over seven verifier languages: the model explains a formal model in natural language with notation comments and a scheme summary, scored 0 to 100 from embedding similarity to verified references and the executability of the annotated code, averaged over languages. Higher is better.
Metric: Interpretation score (0-100) on CrypFormBench (700 instances over 677 cryptographic schemes and 7 formal verifier languages: Scyther, Tamarin, AVISPA, ProVerif, Maude-NPA, CryptoVerif, EasyCrypt): explain a formal model in natural language (notation-level comments and a global scheme summary), scored 0.3 logic-description similarity + 0.3 annotation similarity (Qwen3-Embedding-8B cosine similarity to the verified references) + 0.4 the executability-based score of the annotated code, averaged over languages; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o | 94.1 | #741 |
| 2 | GPT-4o Mini | 93.8 | #1623 |
| 3 | DeepSeek R1 | 77.4 | #473 |
| 4 | Gemini 2.5 Pro | 69.1 | #266 |
| 5 | GLM-4 | 62.3 | #1666 |
| 6 | Grok 3 | 49.6 | #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-interpretation · How It Works · Data refreshed daily, snapshot 2026-11-13.