Those routing decisions have patterns.
Original titleThose routing decisions have patterns. For instance, in one layer, just 64 of 8,128 possible expert pairs accounted for 42% of tokens. I...
AISummary
For instance, in one layer, just 64 of 8,128 possible expert pairs accounted for 42% of tokens. If experts were picked independently, those pairs would carry 1.6%. We show how to exploit this structure to improve where experts are placed.
Source: Zyphra · x.com