Search papers, labs, and topics across Lattice.
VoxReason introduces a novel listener-free evaluation framework for assessing the source-grounded speech planning of expressive speech systems prior to waveform synthesis. By measuring the legality of delivery choices against cited source records, the framework reveals critical insights into how well these systems adhere to source material, highlighting significant discrepancies in citation accuracy. The study demonstrates that a 7B model can significantly improve plan-slot accuracy and locality when trained with a locality-focused repair approach, underscoring the limitations of traditional accuracy metrics in evaluating source-grounded speech systems.
A 7B model can boost source-grounded speech planning accuracy from 68.4% to 91.9% by addressing citation legality before synthesis.
Expressive speech systems make a decision before any waveform is rendered: how an utterance is delivered. In dialogue agents, narration, and role-conditioned TTS, that hidden planning step sets affect, pitch, energy, rate, pause, emphasis, and stance, yet downstream audio scores rarely reveal whether those choices were licensed by the source record, a source-use failure that occurs before any waveform exists. VoxReason makes that pre-synthesis decision measurable as a listener-free task for source-grounded speech planning. Before synthesis, VoxReason measures whether delivery choices are grounded in cited source records. Systems output a source-cited speaking-plan with evidence citations, and a deterministic verifier checks citation legality, slot agreement, unsupported state, schema validity, and one-cue counterfactual locality. On 1,440 checked source-label cases, shortcut controls show why slot accuracy alone is unsafe: a key-lookup oracle reaches 1.000 plan-slot accuracy on seen keys, while an emotion prior still reaches 0.958 slot accuracy on source-key-disjoint cases without citing intensity or identity. In a separate 100-case learned source-key-disjoint comparison, a 7B locality SFT+CF repair improves plan-slot accuracy/locality from 0.684/0.141 to 0.919/1.000, and removing source records lowers citation-required grounded score by 0.488. Rendered waveform quality remains outside the present evaluation.
