Optimal Batch Size Scaling on Li dataset 2025 (train val)
0.667Scaling Parameter3TL
Evaluation Results
| Method | Links | |
|---|---|---|
| 3TLComment=From (4), dataset=Li dataset2026.07 | 0.667 | |
| Step-LawReference=(Li et al., 2025a)2026.07 | 0.58 | |
| DeepSeek-AI et al.Reference=(DeepSeek-AI et al., 2024), Context=From M⋆ ∝ C0.327, D⋆ ∝ C0.4752026.07 | 0.086 | |
| Bergsma et al.Reference=(Bergsma et al., 2025)2026.07 | 0.0306 |