LiME achieves expert specialization in MoE-PEFT via lightweight output modulation on a shared PEFT module plus zero-parameter routing, delivering competitive performance with up to 4x fewer trainable parameters and 29% faster training on the MMT-47 multimodal multi-task benchmark.
All datasets are formatted as text generation tasks with consistent prompt templates
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
LiME: Lightweight Mixture of Experts for Efficient Multimodal Multi-task Learning
LiME achieves expert specialization in MoE-PEFT via lightweight output modulation on a shared PEFT module plus zero-parameter routing, delivering competitive performance with up to 4x fewer trainable parameters and 29% faster training on the MMT-47 multimodal multi-task benchmark.