PROST-LLM: Progressively Enhancing the Speech-to-Speech Translation Capability in LLMs

Jing Xu, Jiaqi Wang, Daxin Tan, Xiao Chen · 2026

Read PROST-LLM: Progressively Enhancing the Speech-to-Speech Translation Capability in LLMs by Jing Xu, Jiaqi Wang, Daxin Tan, Xiao Chen (2026) 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-2601-16618.txt.

Description

Jing Xu and three coauthors grow speech to speech skills in large models step by step. Their 2026 preprint, named PROST-LLM, deals with direct speech swaps. Speech to speech keeps tone and feel that text paths lose. Large models shine on text tasks but lag on speech swaps. Scant speech swap data is a main cause of that lag.

Step growth adds skills in a set order to beat scant data. Each step builds on the last with tasks of more depth. Gains would open speech swaps to more tongue pairs. They would also cut need for rare paired speech sets. The work gives a staged path for that growth.

Catalog record

CREATORJing Xu, Jiaqi Wang, Daxin Tan, Xiao Chen
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE22 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-16618

Browse the full library · About the Tower of Babel