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

General Capability Retention research benchmarks · SOTA2 Research

General Capability Retention

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

Held-out capability average

82.6Unweighted Average Accuracy

21

May 19, 2026

GSM8K, MMLU, ARC-C

68.5GSM8K

14

Feb 26, 2026

Popular datasets

GSM8K, MMLU, ARC Held-out capability

Follow for updates

Join SOTA2 Contact