CryptanalysisBench provides a standardized evaluation framework for measuring LLM performance on cryptographic tasks. Developed with ETH Zurich and Tel Aviv University, it quantifies how effectively models approach code breaking and mathematical pattern recognition.
HOW THIS AFFECTS YOU
●
researcherYou can use this benchmark to assess model vulnerability to cryptographic reasoning.
●
policyThis helps you understand the security risks posed by advanced cryptographic capabilities in LLMs.