card your ML experiment.

Describe an ML experiment — dataset, version, splits, seed, hyperparameters, training, framework, compute, evaluation, and what you released. MLReproCard scores a real weighted reproducibility rubric (grounded in the NeurIPS/ICML reproducibility checklists, Mitchell et al. 2019 Model Cards, and Gundersen's reproducibility taxonomy) across data, code, training, evaluation, compute, and sharing — flags exactly which repro elements are missing, assigns an R0–R3 level and an overall 0–100 score, and fills in a normalized model card. Deterministic; no LLM, no network.