Language Model Evaluation Suite on LM Evaluation Harness
66.6Avg AccuracySpanNorm
Evaluation Results
| Method | Links | |
|---|---|---|
| SpanNormParam=5B, Tokens=200B, Architecture=Dense2026.01 | 66.6 | |
| SpanNormParam=A2.4B-16B, Tokens=200B, Architecture=MoE2026.01 | 66.3 | |
| PreNormParam=A2.4B-16B, Tokens=200B, Architecture=MoE2026.01 | 65.1 | |
| PreNormParam=5B, Tokens=200B, Architecture=Dense2026.01 | 64.1 | |
| SpanNormParam=1.3B, Tokens=100B, Architecture=Dense2026.01 | 58 | |
| PreNormParam=1.3B, Tokens=100B, Architecture=Dense2026.01 | 55.8 | |
| SpanNormParam=740M, Tokens=30B, Architecture=Dense2026.01 | 50.8 | |
| PreNormParam=740M, Tokens=30B, Architecture=Dense2026.01 | 49.8 |