NLE: Non-autoregressive LLM-based ASR by Transcript Editing

Avihu Dekel; Samuel Thomas; Takashi Fukada; George Saon

返回文献探索

Paper · arXiv 2603.08397

NLE: Non-autoregressive LLM-based ASR by Transcript Editing

Avihu Dekel, Samuel Thomas, Takashi Fukada, George Saon

24 upvotesMarch 9, 2026arXiv 预印本

AI 摘要

A non-autoregressive speech recognition approach formulates acoustic-to-text conversion as conditional transcript editing using a bidirectional language model editor with latent alignment training and interleaved padding for improved efficiency.

autoregressivenon-autoregressivespeech recognitionconditional transcript editingacoustic embeddingspretrained speech encoderbidirectional LLM editorlatent alignment objectiveinterleaved paddingTransformerreal-time factorWER

Abstract

While autoregressive (AR) LLM-based ASR systems achieve strong accuracy, their sequential decoding limits parallelism and incurs high latency. We propose NLE, a non-autoregressive (NAR) approach that formulates speech recognition as conditional transcript editing, enabling fully parallel prediction. NLE extracts acoustic embeddings and an initial hypothesis from a pretrained speech encoder, then refines the hypothesis using a bidirectional LLM editor trained with a latent alignment objective. An interleaved padding strategy exploits the identity mapping bias of Transformers, allowing the model to focus on corrections rather than full reconstruction. On the Open ASR leaderboard, NLE++ achieves 5.67% average WER with an RTFx (inverse real-time factor) of 1630. In single-utterance scenarios, NLE achieves 27x speedup over the AR baseline, making it suitable for real-time applications.