Computer Science · Artificial Intelligence
Among 200 high school students in Shiraz, Iran, open family communication and some forms of constructive inner self-talk were associated with lower reported cyberbullying victimization. Family communication focused on conformity was associated with more victimization, but the study cannot show that these factors cause or prevent cyberbullying.
Researchers developed an AI system that identifies offensive Arabic social-media text while using significantly fewer fine-tuning samples than existing methods. The system combines related language tasks, uncertainty-based sample selection and weighted emoji handling.
The European Union’s AI Act calls for providers to mark and detect outputs from general-purpose AI models. A new analysis finds that no current watermarking approach for large language models meets all four of the law’s stated standards: reliability, interoperability, effectiveness and robustness.
A preprint study found that the same hidden instruction could produce very different results in AI agents depending on whether it arrived as tool information or tool output, and what surrounded it. The findings have not been peer reviewed and may change.
