● REPRODUCIBLE RELEASERP
RESEARCH INFRASTRUCTURE / MODEL GOVERNANCE
ModelGuard
Observatory.
Evidence-based evaluation monitoring for multilingual NLP. Examine model performance, dataset shift, class-level errors and reproducible release criteria.
4 deterministic models3 languages6 rhetorical classeszero-server compute
ARTIFACT PROVENANCE● VERIFIED LOCALLY
736Prediction records
2Evaluation splits
842Training records
4Baseline methods
Academic Rhetorical Intent Mini v1.1 · curated synthetic sentences. Evaluation results are not evidence of real-world deployment reliability.
Loading signed evaluation artifacts…