Embracing Diversity: A Multi-Perspective Approach with Soft Labels
Benedetta Muscato *, Praveen Bushipaka *, Gizem Gezici, Lucia Passaro, Fosca Giannotti, Tommaso Cucinotta · 2025
Read Embracing Diversity: A Multi-Perspective Approach with Soft Labels by Benedetta Muscato *, Praveen Bushipaka *, Gizem Gezici, Lucia Passaro, Fosca Giannotti, Tommaso Cucinotta (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-2503-00489.txt.
Description
Benedetta Muscato and five co-authors argue that when people disagree on labels, the disagreement itself is data. Their 2025 paper works on stance detection, the task of judging where a text stands on a disputed topic. Human judges from unlike backgrounds often split on such calls, and the usual fix of one gold label erases that split.
Their framework trains models on soft labels that keep the spread of views, and tests them against that spread too. They share a new data set labeled by both humans and large language models. Models trained the new way scored higher on standard accuracy measures than single-label models. They also showed lower confidence, which the authors link to the task's built-in subjectivity. The work aims at fairer, more honest language tools.
Catalog record
| CREATOR | Benedetta Muscato *, Praveen Bushipaka *, Gizem Gezici, Lucia Passaro, Fosca Giannotti, Tommaso Cucinotta |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-00489