docs: clarify Marimba model usage, outputs, and evaluation
#1
by todeschini - opened
Restructure the public model card around a runnable synthetic-audio quickstart, typed outputs, existing evaluation results, limitations, licensing, and a provisional paper citation. Preserve all reported results and Hub metadata. Reuse the existing repository-hosted Marimba brand assets unchanged.
No model weights, configuration, benchmarks, or runtime code change.
Reviewed: metrics unchanged, quickstart verified, no private data.
todeschini changed pull request status to merged