GASTON: Graph-Aware Social Transformer for Online Networks
Olha Wloch, Liam Hebert, Robin Cohen, Lukasz Golab · 2026
Read GASTON: Graph-Aware Social Transformer for Online Networks by Olha Wloch, Liam Hebert, Robin Cohen, Lukasz Golab (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-2602-02524.txt.
Description
Olha Wloch and 3 co-authors argue that harmful content online cannot be judged from words alone, in a 2026 arXiv paper. The same sentence can be support in one community and abuse in another. Meaning depends on local norms, so detection models need community context built in.
Their system, GASTON, learns text and user embeddings grounded in those local norms. A contrastive pretraining step builds community embeddings from user membership patterns before any text is read. That lets the system tell a support group from a hate group by who gathers there, even when vocabularies overlap. On stress detection, toxicity scoring, and norm violation tasks, the embeddings beat state-of-the-art baselines. The design treats community identity as first-class evidence for moderation.
Catalog record
| CREATOR | Olha Wloch, Liam Hebert, Robin Cohen, Lukasz Golab |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 57 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-02524