Report
Claimed stabilization of MoE models at 256x sparsity
A quote post pairs “0,39%” with a claim about stabilizing mixture-of-experts models at 256x sparsity.
TLDR
A quote post highlights “0,39%” alongside a claim about stabilizing mixture-of-experts (MoE) models at 256x sparsity. It does not say what the percentage measures.
Combined views
1.6K
1 Source, first seen 2h ago
10 likes