Decoding Strategies for Diffusion-Based ASR: A Systematic Evaluation of Confidence-Based Thresholding

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yeo, Jeong Hun, Kim, Minsu, Rha, Hyeongseop, Ro, Yong Man
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917543298465792
author Yeo, Jeong Hun
Kim, Minsu
Rha, Hyeongseop
Ro, Yong Man
author_facet Yeo, Jeong Hun
Kim, Minsu
Rha, Hyeongseop
Ro, Yong Man
contents While LLM-based Automatic Speech Recognition (ASR) achieves high accuracy, its speed is limited by sequential autoregressive decoding. Diffusion Language Models (DLMs) offer a parallel alternative, yet their decoding strategies remain under-explored in ASR contexts. This paper analyzes three decoding schemes for DLM-based ASR: fixed-number, static confidence threshold, and dynamic confidence threshold. We propose measuring round-wise accuracy using Negative Log-Likelihood-based uncertainty as a proxy for decoding progress. Our results show that both threshold-based strategies significantly outperform fixed-number schemes in accuracy and speed. We attribute this to a property unique to ASR: most tokens reach high confidence early, allowing reliable ones to be harvested aggressively while leaving only difficult tokens for later rounds. Notably, the static-threshold strategy matches the accuracy of autoregressive decoding while offering superior efficiency.
format Preprint
id arxiv_https___arxiv_org_abs_2605_29613
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Decoding Strategies for Diffusion-Based ASR: A Systematic Evaluation of Confidence-Based Thresholding
Yeo, Jeong Hun
Kim, Minsu
Rha, Hyeongseop
Ro, Yong Man
Audio and Speech Processing
Sound
While LLM-based Automatic Speech Recognition (ASR) achieves high accuracy, its speed is limited by sequential autoregressive decoding. Diffusion Language Models (DLMs) offer a parallel alternative, yet their decoding strategies remain under-explored in ASR contexts. This paper analyzes three decoding schemes for DLM-based ASR: fixed-number, static confidence threshold, and dynamic confidence threshold. We propose measuring round-wise accuracy using Negative Log-Likelihood-based uncertainty as a proxy for decoding progress. Our results show that both threshold-based strategies significantly outperform fixed-number schemes in accuracy and speed. We attribute this to a property unique to ASR: most tokens reach high confidence early, allowing reliable ones to be harvested aggressively while leaving only difficult tokens for later rounds. Notably, the static-threshold strategy matches the accuracy of autoregressive decoding while offering superior efficiency.
title Decoding Strategies for Diffusion-Based ASR: A Systematic Evaluation of Confidence-Based Thresholding
topic Audio and Speech Processing
Sound
url https://arxiv.org/abs/2605.29613