List your company

Language Modeling on Tülü 3 SFT Mixture

9.96Perplexity (PPL)4-vium
9.474812.749916.02519.3001May 13, 2026

Evaluation results

MethodLinks
4-viumDepth=48, Hidden width=768, beta=0.152026.059.9618.45
4-viumDepth=24, Hidden width=768, beta=0.152026.0510.9121.73
4-viumDepth=36, Hidden width=768, beta=0.152026.0511.479.15
BaselineDepth=48, Hidden width=7682026.0512.22—
BaselineDepth=36, Hidden width=7682026.0512.63—
4-viumDepth=12, Hidden width=768, beta=0.152026.0513.9314.78
BaselineDepth=24, Hidden width=7682026.0513.94—
BaselineDepth=12, Hidden width=7682026.0516.35—
4-viumDepth=6, Hidden width=768, beta=0.152026.0517.3721.34
BaselineDepth=6, Hidden width=7682026.0522.09—