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

Large Model Performance Prediction research benchmarks · SOTA2 Research

Large Model Performance Prediction

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

Large Model Performance Prediction dataset 1.0 (40% masking)

6.13RMSE

10

Feb 26, 2026

Large Model Performance Prediction 60% masking

6.77RMSE

10

Feb 26, 2026

OpenCompass 95% masking September 30, 2024 cutoff (temporal split)

8.75RMSE

10

Feb 26, 2026

Benchmark-side Pattern Shift Math

46.59Average Score

6

Feb 26, 2026

285 models on one Math benchmark

Brute-force Evaluation

100Top-10 Recall

5

Feb 26, 2026

Benchmark Chinese pattern shift

16.94RMSE

3

Feb 26, 2026

Benchmark OCR pattern shift

25.18RMSE

3

Feb 26, 2026

Frontier Top-20 pattern shift

9.71RMSE

3

Feb 26, 2026

Paradigm RLHF pattern shift

9.55RMSE

3

Feb 26, 2026

Architecture pattern shift MoE

10.68RMSE

3

Feb 26, 2026

Benchmark-side Pattern Shift Chinese

42.16Average Score

3

Feb 26, 2026

Benchmark-side Pattern Shift OCR

47.6Score Avg

3

Feb 26, 2026

Model-side Pattern Shift Frontier

13.22Score Avg

3

Feb 26, 2026

Model-side Pattern Shift Paradigm

11.74Score Avg

3

Feb 26, 2026

Model-side Pattern Shift Architecture

10.64Avg Score

3

Feb 26, 2026

Popular datasets

Model-side Pattern Shift Benchmark-side Pattern Shift benchmark MATH OpenCompass

Follow for updates

Join SOTA2 Contact