Spoken Language Understanding on Unseen Tasks With In-Context Learning

Neeraj, Sriram · 2025

Read Spoken Language Understanding on Unseen Tasks With In-Context Learning by Neeraj, Sriram (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-2505-07731.txt.

Description

Neeraj Agrawal and Sriram Ganapathy teach speech-understanding models to handle tasks they have never seen, in this 2025 arXiv paper. Spoken language tasks range from pulling facts out of speech to sorting intent, and training data for each new one is often missing. Speech-text language models promise a way out, yet their out-of-the-box scores on such tasks disappoint.

The authors fine-tune with scrambled class labels, forcing the model to learn the shape of a task from a few examples in the prompt rather than memorize fixed answers. On unseen tasks, the trick lifted performance well past standard fine-tuning and plain prompting. No fresh annotation was needed to switch the model to a new task.

Catalog record

CREATORNeeraj, Sriram
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE21 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-07731

Browse the full library · About the Tower of Babel