01 / 2 ARTICLES
Evaluation foundations
Preserve the run record and inspect the inputs before interpreting outputs.
A reproducible model evaluation starts with a run record →MODEL PRACTICE / 03
Make the workflow.
Measure what happens.
Protocols and practical guides for reproducible AI model evaluation, from audio to multimodal workflows.
Explore the libraryExplore by question
01 / FIELD NOTES02 / 1 ARTICLE
Audio protocols
A practical starting procedure for speech-generation comparisons.
A practical protocol for comparing text-to-speech outputs →From the notebook
02 / RECENT ENTRIESCheck multimodal inputs before comparing model outputs
A preflight protocol for image, audio, video, and text evaluation inputs.
A reproducible model evaluation starts with a run record
What to save so another person can understand and repeat a model evaluation.
A practical protocol for comparing text-to-speech outputs
A proposed evaluation procedure for pronunciation, intelligibility, voice behavior, and timing.
A GROWING REFERENCE
Across model types. Grounded in sources.
Our scope includes language, vision, video, audio, multimodal systems, and emerging world models. The library begins with the foundations and audio evaluation; new coverage follows available evidence.