A supervised mixture of experts, routed by fixed bandwidth and task labels, improves multi-task ASR/ST over hard parameter sharing while keeping active parameters constant.
Decoder S-MoE To validate the effectiveness of S-MoE applied to the decoder blocks, we conducted experiments on two tasks: Korean-to- English (ko2en) ST and Korean ASR (ko-ASR)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Beyond Hard Sharing: Efficient Multi-Task Speech-to-Text Modeling with Supervised Mixture of Experts
A supervised mixture of experts, routed by fixed bandwidth and task labels, improves multi-task ASR/ST over hard parameter sharing while keeping active parameters constant.