Paradigma's 1B math model: sample-efficient recipe and open evals
- Source
- x.com
- Date
We are releasing the technical report for Limite 1B -Violetto. Violetto is our first model, designed to push mathematical capabilities per parameter to its limits, trained from scratch.
Context
Paradigma's Limite 1B - Violetto is a one-billion-parameter model specialized for solving one mathematical problem at a time. Its model card recommends supplying the current question without conversation history and warns that it can reinterpret unrelated requests as mathematical tasks. That specialization determines what developers should test it for.
The release provides a way to check its mathematical performance: a public evaluation package fixes model versions and prompt templates, while the repository documentation explains how runs record dataset versions, sampling settings and scores. These records let evaluators inspect the conditions behind a result and rerun the tests.
- eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
- benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
Checking sign-in…
Loading comments…