Reproduce “JSTprove: Pioneering Verifiable AI for a Trustless Future” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Jonathan Gold, Tristan Freiberg, Haruna Isah, Shirin Shahabi

JSTprove provides an ONNX-to-proof workflow for checking a model’s inference. Its command-line interface coordinates quantization, circuit compilation, witness creation, proof generation, and verification. The authors demonstrate it on a LeNet-style convolutional neural network. Mean circuit compilation time rises from 706.60 seconds at 56×56 pixels to 2,657.29 seconds at 84×84; compilation at 112×112 runs out of memory on their 36-GB machine.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.