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

Speech Recognition research benchmarks – Page 5 · SOTA2 Research

Speech Recognition

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

Turkish video dataset (noisy)

+ crosslingual pretrain + aux + kl loss

18.1WER

9

Feb 26, 2026

Turkish video dataset (clean)

+ crosslingual pretrain + aux + kl loss

16.1WER

9

Feb 26, 2026

Switchboard Eval2000

9SW Error Rate

9

Feb 26, 2026

LibriSpeech clean 1000h (test)

0.038WER

9

Feb 26, 2026

Aura-STT-BenchLF (DA)

Gemini 2.5 Flash

38.48WER

8

Mar 18, 2026

Aura-STT-BenchLF MSA

16.3WER (%)

8

Mar 18, 2026

11WER

8

Feb 26, 2026

LibriSpeech Clean

0.013WER

8

Feb 26, 2026

Libri2Mix max mode (test)

15.6WER (%)

8

Feb 26, 2026

14.6WER

8

Feb 26, 2026

AF-Next-Instruct

5.4WER

7

Apr 14, 2026

BABEL Swahili sw (test)

21WER

7

Feb 26, 2026

BABEL Tagalog tl (test)

29.3WER

7

Feb 26, 2026

0.209PER

7

Feb 26, 2026

Dutch dysarthric speech (spontaneous)

34.4Word Error Rate (WER)

6

Jun 30, 2026

Dutch dysarthric speech (customized read)

23.7WER

6

Jun 30, 2026

Dutch dysarthric speech (generic read)

26.1WER

6

Jun 30, 2026

TDRC + nonlinear Duffing oscillator + Lyon cochleagram

99.8Accuracy

6

May 29, 2026

45.4WER (%)

6

Mar 20, 2026

AMI IHM (unlabeled)

10.09WER

6

Feb 26, 2026

9.35WER (%)

6

Feb 26, 2026

SPGISpeech (unlabeled)

2.23WER

6

Feb 26, 2026

Earnings-22 (unlabeled)

10.63WER (%)

6

Feb 26, 2026

Earnings-22 (test)

11.82WER

6

Feb 26, 2026

Earnings 21 (unlabeled)

7.79WER

6

Feb 26, 2026

Page 5 of 9

Popular datasets

LibriSpeech Haitan internal CHiME Haitan Vibration Sensor Audio

Follow for updates

Join SOTA2 Contact