List your company
ResearchFormerly WizWand

Knowledge on MMLU-Redux (test)

69.8Knowledge AccuracyReM-MoA*
21.887234.326146.76559.2039Jun 13, 2025Aug 14, 2025Oct 16, 2025Dec 17, 2025Feb 18, 2026Apr 21, 2026Jun 23, 2026
Updated 4mo ago

Evaluation results

MethodLinks
ReM-MoA*Width (N)=8, Depth (L)=3, Trend=↑↑, Distilled Reviewer=true2026.0669.8
ReM-MoA*Width (N)=7, Depth (L)=3, Trend=↑↑, Distilled Reviewer=true2026.0669.5
ReM-MoA*Width (N)=6, Depth (L)=3, Trend=↑↑, Distilled Reviewer=true2026.0669.3
ReM-MoAWidth (N)=8, Depth (L)=3, Trend=↑→2026.0669.2
ReM-MoAWidth (N)=7, Depth (L)=3, Trend=↑→2026.0669
ReM-MoAWidth (N)=6, Depth (L)=3, Trend=↑→2026.0668.8
ReM-MoA*Width (N)=5, Depth (L)=3, Trend=↑↑, Distilled Reviewer=true2026.0668.7
ReM-MoAWidth (N)=5, Depth (L)=3, Trend=↑→2026.0668.3
ReM-MoA*Width (N)=4, Depth (L)=3, Trend=↑↑, Distilled Reviewer=true2026.0668.2
ReM-MoAWidth (N)=4, Depth (L)=3, Trend=↑→2026.0667.8
ReM-MoA*Width (N)=3, Depth (L)=3, Trend=↑↑, Distilled Reviewer=true2026.0667.4
ReM-MoAWidth (N)=3, Depth (L)=3, Trend=↑→2026.0667
RMoAWidth (N)=4, Depth (L)=3, Trend=→2026.0665.5
RMoAWidth (N)=5, Depth (L)=3, Trend=→2026.0665.5
RMoAWidth (N)=3, Depth (L)=3, Trend=→2026.0665.4
RMoAWidth (N)=6, Depth (L)=3, Trend=→2026.0665.4
RMoAWidth (N)=7, Depth (L)=3, Trend=→2026.0665.3
RMoAWidth (N)=8, Depth (L)=3, Trend=→2026.0665.2
AttentionMoAWidth (N)=5, Depth (L)=3, Trend=→2026.0665.2
AttentionMoAWidth (N)=6, Depth (L)=3, Trend=→2026.0665.2
AttentionMoAWidth (N)=4, Depth (L)=3, Trend=→2026.0665.1
AttentionMoAWidth (N)=7, Depth (L)=3, Trend=→2026.0665.1
AttentionMoAWidth (N)=8, Depth (L)=3, Trend=→2026.0665
AttentionMoAWidth (N)=3, Depth (L)=3, Trend=→2026.0664.8
Standard MoAWidth (N)=3, Depth (L)=3, Trend=↓2026.0658.2
Standard MoAWidth (N)=4, Depth (L)=3, Trend=↓2026.0657.2
Standard MoAWidth (N)=5, Depth (L)=3, Trend=↓2026.0656.5
Standard MoAWidth (N)=6, Depth (L)=3, Trend=↓2026.0655.8
Standard MoAWidth (N)=7, Depth (L)=3, Trend=↓2026.0655.2
Standard MoAWidth (N)=8, Depth (L)=3, Trend=↓2026.0654.5
MoE w/ optimal ARModel Type=MoE, Activation rate=20.07, Compute=2.86e21, Data reuse=-2025.0630.93
Dense baselineModel Type=Dense, Compute=5.45e212025.0628.9
MoE w/ optimal ARModel Type=MoE, Activation rate=20.07, Compute=2.86e21, Data reuse=strict2025.0623.73