The CCF AATC 2025 Speech Restoration Challenge: A Retrospective

Junan Zhang, Mengyao Zhu, Xin Xu, Hui Bu, Zhenhua Ling, Zhizheng Wu · 2025

Read The CCF AATC 2025 Speech Restoration Challenge: A Retrospective by Junan Zhang, Mengyao Zhu, Xin Xu, Hui Bu, Zhenhua Ling, Zhizheng Wu (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-12974.txt.

Description

Junan Zhang and 5 co-authors look back on a 2025 speech repair contest in this paper. Real speech is rarely hurt in just one way. Noise in the room, tight codec squeezing, and flaws left by past repair tools all mix, and lab tests often miss that mix.

The contest asked for one model that can repair speech blind, with no hint of what went wrong. It had to cope with three kinds of harm: room noise and the like, codec harm, and flaws left by past tools. The paper tells how the data set was built and how the task was set, and it studies the twenty-five systems that took part. One key find stood out: small, lean models with few weights beat the trend toward huge generative models.

Catalog record

CREATORJunan Zhang, Mengyao Zhu, Xin Xu, Hui Bu, Zhenhua Ling, Zhizheng Wu
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE41 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-12974

Browse the full library · About the Tower of Babel