What do Speech Foundation Models Learn? Analysis and Applications

Ankita Pasad · 2025

Read What do Speech Foundation Models Learn? Analysis and Applications by Ankita Pasad (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-2508-12255.txt.

Introduction

Ankita Pasad wrote a 2025 thesis on what speech models learn from sound. Speech foundation models aim to serve many tasks, from typing speech to finding names in speech. New models keep scoring well, yet what they store in each layer is less clear.

Pasad builds a light test frame that needs no fresh training. It probes both sound traits and word traits across layers of many models. The tests show that layer traits can predict how well a model does on real tasks. The thesis also adds new tasks for finding names in spoken speech, a field held back by few data sets.

Catalog record

CREATORAnkita Pasad
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE253 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-12255

Browse the full library · About the Tower of Babel