UniSRM is a unified speech reward model with new datasets that uses a two-stage reasoning pipeline to deliver interpretable, human-aligned evaluations across utterance quality to context coherence tasks.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
UniSRM: A Unified Speech Reward Model for Reasoning-Based Fine-grained Assessment
UniSRM is a unified speech reward model with new datasets that uses a two-stage reasoning pipeline to deliver interpretable, human-aligned evaluations across utterance quality to context coherence tasks.