List your company

Hyper-parameter Search on Dense-8B

1.02FLOPsGrid Search
0.73482.65994.5856.5101Jun 4, 2026

Evaluation results

MethodLinks
Grid SearchStrategy=Grid Search, Proxy Models=9 × Dense-8B2026.061.02—
Ours (Scaling)Strategy=Ours (Scaling), Proxy Models=9 × {100M, 500M}2026.068.1592