Meta Expert: cross-layer expert sharing matches MoE with 70-80% experts

Tsinghua and Shanghai AI Lab propose Meta Expert, a cross-layer expert-sharing MoE architecture that matches mainstream models' performance using only 70-80% of the experts, reducing cross-layer redundancy.

2026-10-22 ~ 2026-10-23 · 3 related posts

• Meta Expert paper: cross-layer shared experts match vanilla MoE with only 70-80% of experts — Xianbao_QIAN · 2026-10-22
• Meta Expert: cross-layer expert sharing matches vanilla MoE with 70-80% experts — pmttyji · 2026-10-23

1 near-duplicate retellings: TsinghuaC3I