Scaling Law Estimation on Scaling Law Analysis Scaling coefficients
0.644Alpha M CoefficientNie et al. (2025a)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Nie et al. (2025a)Model type=Autoregressive2025.12 | 0.644 | 0.356 | -0.0633 | |
| MDM (Nie et al., 2025a)Model type=Diffusion2025.12 | 0.634 | 0.366 | -0.0615 | |
| DLM (uniform)Model type=diffusion, Noise type=uniform2025.12 | 0.589 | 0.411 | -0.0522 | |
| DLM (high-uniform)Model type=diffusion, Noise type=high-uniform2025.12 | 0.573 | 0.427 | -0.0514 | |
| DLM (masked)Model type=diffusion, Noise type=masked2025.12 | 0.566 | 0.434 | -0.0496 | |
| DLM (low-uniform)Model type=diffusion, Noise type=low-uniform2025.12 | 0.535 | 0.465 | -0.0509 | |
| DLM (balanced)Model type=diffusion, Noise type=balanced2025.12 | 0.534 | 0.466 | -0.0512 | |
| Bi et al. (2024)Model type=Autoregressive2025.12 | 0.5243 | 0.4757 | — | |
| MDM (Ni et al., 2025)Model type=Diffusion2025.12 | 0.514 | 0.486 | — | |
| Hoffmann et al. (2022)Model type=Autoregressive2025.12 | 0.49 | 0.51 | — | |
| Shuai et al. (2024)Model type=Autoregressive2025.12 | 0.464 | 0.536 | — |