Get started with Witnesses: generate dataset provenance certificates in PyTorch or JAX
pratyusha_PS · x · 2026-11-07
Stanford CS master's student Houjun Liu and Pratyusha Sharma invite researchers to try Witnesses on their next training run. The DVBench tutorial shows how to attest what datasets a training job used:
- pip install "dvbench[torch]" or "dvbench[jax]"; import dvbench.Client before Torch/JAX and connect to the production API
- Register an experiment, initialize a zkdv proof tape, and configure name (kebab-case) and rollingwindow (tokens per sample)
- Flax TrainState's applygradients() workflow is supported via a separate guide; leaderboards available
A practical tool for auditable training-data provenance claims.
Related event: Witnesses Lets AI Training Claims Be Verified Without Reruns(2 posts)→
More from coding & agent
- CodeRabbit podcast: ship less, not more, in the AI era — nutlope · 2026-11-07
- 16 MCP servers, 35 parallel agents — and it still ships slop — weswinder · 2026-11-07
- Creator apps like Instagram Edits and Roblox Studio are shipping with built-in MCPs for agents — trq212 · 2026-11-07
- Cartesian agent builds a 289-part Dymaxion car 3D model live, researching the design online — carlosbannon · 2026-11-07
- ArkenOS visualizes hundreds of autonomous agents collaborating as a 3D research swarm — _AustinCalvert_ · 2026-11-07
- AI agents blocked by Zillow, Indeed, Reddit — a scraping API got 3/3 through in testing — eyishazyer · 2026-11-07