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

Multi-task Generalization research benchmarks · SOTA2 Research

Multi-task Generalization

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

Other tasks (test)

60.28Score

36

Feb 26, 2026

Multi-task Overall Average

40.5Accuracy

11

May 29, 2026

Countdown and OOD Tasks Overall (test)

R1 Distill -> GRPO

35.9Accuracy

10

Feb 26, 2026

StoryCloze, OpenQA, ARC-E, ARC-C combined

87.76Average Accuracy

8

Mar 4, 2026

Real Hardware 4 tasks (untrained)

9Stage 1 Success Rate

4

May 19, 2026

Real Hardware 8 tasks (train)

0.875Stage 1 Success Rate

4

May 19, 2026

Popular datasets

Real Hardware Other tasks Countdown and OOD Tasks StoryCloze, OpenQA, ARC Multi-task

Follow for updates

Join SOTA2 Contact