ResearchBenchmarksAutomatic Speech Recognition on LibriSpeech clean 10h (dev)Automatic Speech Recognition on LibriSpeech clean 10h (dev)Follow0.2834Word Error RateAIPA0.28060.29950.31840.3373Jun 22, 2024Evaluation resultsMethodMethodLinksWord Error RateAIPAArchitecture=Enc-CTC,...Architecture=Enc-CTC, Data Augmentation=SpecAugment + AIPA, Auxiliary Loss=InterCTC, Technique=COS2024.060.2834BaselineArchitecture=Enc-CTC,...Architecture=Enc-CTC, Data Augmentation=SpecAugment, Auxiliary Loss=InterCTC2024.060.3534