Categories Companies Research

Loading the SOTA2 catalog…

Explore

Categories Companies Research

Contribute

Join SOTA2 Contact

SOTA2

© 2026 SOTA2

Build in Bellevue, USA

WizWand is now SOTA2 Research

Papers Benchmarks Datasets Tasks

Scaling Law Fitting research benchmarks · SOTA2 Research

Scaling Law Fitting

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

NMM interleaved (test)

2.693MRE

23

May 19, 2026

NMM-captions (test)

0.037MRE

23

May 19, 2026

LVM highquality 2 (test)

0.075Mean Relative Error (MRE)

23

May 19, 2026

LVM-highquality 1 (test)

0.076MRE

23

May 19, 2026

LVM-alttext (test)

3.083MRE

23

May 19, 2026

LLM-wikipedia (test)

0.085MRE

23

May 19, 2026

LLM-stackexchange (test)

6.109MRE

23

May 19, 2026

LLM-github (test)

8.597MRE

23

May 19, 2026

LLM-commoncrawl (test)

0.04Mean Relative Error

23

May 19, 2026

0.041MRE

23

May 19, 2026

LLM-book (test)

3.671Mean Relative Error (MRE)

23

May 19, 2026

LLM-arxiv (test)

6.848MRE

23

May 19, 2026

NMM text (test)

0.023MRE

14

May 19, 2026

LVM-synthetic (test)

9.318MRE

14

May 19, 2026

Perturbation-aware Baseline (Exponential)

0.98Performance (1e-5)

7

May 25, 2026

Shannon Scaling Law

0.987Score at Scaling Factor 1e-5

7

May 25, 2026

Shannon Scaling Law

0.986Performance (Scale 1e-5)

7

May 25, 2026

Shannon Scaling Law

0.989Performance (4-bit)

7

May 25, 2026

Shannon Scaling Law

99.53Performance (4-bit)

7

May 25, 2026

0.405Scaling Parameter b

5

Feb 26, 2026

DCLM strongly regularized baseline 20 points (Full fit)

0.008RMSE

4

Jun 8, 2026

Regularized-baseline points

0.008RMSE

4

Jun 8, 2026

Scaling Law Datasets (Full fit)

0.008RMSE

4

Jun 8, 2026

Baseline Scaling Dataset 400M (Held-out)

0.006RMSE

4

Jun 8, 2026

Baseline Scaling Dataset (Full fit)

0.008RMSE

4

Jun 8, 2026

Popular datasets

NMM GSM8K SIQA Pythia DCLM

Follow for updates

Join SOTA2 Contact