Multi-Router
A submodule of Multi-View MoE that aggregates the sigmoid scores of multiple independently initialized routers. It exploits differentiated preferences to broaden the diversity of candidate signals without increasing the number of activated experts.
Timeline (last 90 days)
Multi-View MoE improves expert selection by aggregating the sigmoid scores of multiple independently initialized routers, expanding the diversity of candidate signals without increasing the number of activated experts.
Using only the Multi-Router submodule alone brings an average score gain of 0.6 percentage points.
All Facts (2)
Multi-View MoE improves expert selection by aggregating the sigmoid scores of multiple independently initialized routers, expanding the diversity of candidate signals without increasing the number of activated experts.
50%UnverifiedUsing only the Multi-Router submodule alone brings an average score gain of 0.6 percentage points.
50%