🔬 Reproducible AI Initiative

Bringing open-source ML models to a fully reproducible state — capturing provenance end to end, registering lineage, and publishing the results (successes and failures) at reproducibleailist.org.

An initiative by TReqs, Inc., dogfooding the TReqs AI stack (roar · GLaaS · TReqs) on real open models.

What you'll find here

• 🧬 Experiments — each reproduction's training run, hosted live.
• 📦 Models — reproduced open models, published with their full provenance.
• 📊 Datasets — mirrored inputs, where licenses permit redistribution.

The method

Every reproduction is run under observation and made independently re-runnable: we record where each input came from and where each output lives, so anyone can verify a result rather than take a claim on faith. Verdicts — green and red alike — are published transparently, and blocked rebuilds are graded as findings about the ecosystem, never about the author.

Learn more at reproducibleailist.org · built with roar and GLaaS.