SERE: Similarity-based Expert Re-routing for Efficient Batch Decoding in MoE Models
Paper
•
2602.07616
•
Published
Jialiang Cheng
Julius-L
·
AI & ML interests
None yet
Recent Activity
authored
a paper
about 20 hours ago
SERE: Similarity-based Expert Re-routing for Efficient Batch Decoding in MoE Models
authored
a paper
about 20 hours ago
EDiT: A Local-SGD-Based Efficient Distributed Training Method for Large Language Models
liked
a dataset
6 months ago
Salesforce/wikitext
Organizations
authored
2
papers
about 20 hours ago