THE LIBRARY
Find your starting point.
Protocols and practical guides for reproducible AI model evaluation, from audio to multimodal workflows.
Evaluation foundations
Preserve the run record and inspect the inputs before interpreting outputs.
- A reproducible model evaluation starts with a run record
What to save so another person can understand and repeat a model evaluation.
- Check multimodal inputs before comparing model outputs
A preflight protocol for image, audio, video, and text evaluation inputs.
Audio protocols
A practical starting procedure for speech-generation comparisons.
- A practical protocol for comparing text-to-speech outputs
A proposed evaluation procedure for pronunciation, intelligibility, voice behavior, and timing.