LLM agents that write machine-checked cryptographic proofs in EasyCrypt (arXiv:2607.02847)

[ECCV 2024 - Oral] Official PyTorch Implementation of "Adversarial Robustification via Text-to-Image Diffusion Models"

Mastering the Art of EasyCrypt Programming

VIGIL — The Provable Adversary · Provable Offensive Security. An autonomous adversary that doesn't ask you to trust it: every finding is minted by a deterministic oracle over the target's bytes — signed, independently replayable. Autonomy behind fail-closed crypto gates; measures its coverage; portable evidence, not reports. Sovereign by design.

🔍 Generate STARK proofs from ZK IR execution traces using a multi-chip architecture optimized for 32-bit integers and the Baby Bear field.