A new substrate for synthetic speech research objects is proposed, which includes generation provenance, source specification, and quality signals, enhancing the ability to attribute model behavior.
> Abstract: Attributing model behavior to synthetic training data requires knowing what produced each training item before estimating what that item caused.
Abstract: Attributing model behavior to synthetic training data requires knowing what produced each training item before estimating what that item caused. A waveform-label pair does not preserve this knowledge. We propose a generation-provenance substrate in which a synthetic research object binds source specification, generated content, waveform, target, fact requirements, quality signals, revie…
Comments: Accepted to the Third NeurIPS Workshop on Attributing Model Behavior at Scale: Data Attribution and Provenance. 4 pages, 0 figures, 1 table.
