Tag: content moderation

An open-source system combining Arabic text analysis, image recognition and uncertainty-based escalation reached a macro F1 score of 0.71 on ...

A prospective audit found that TikTok returned “no violation” for 95.2% of user reports about UK-facing drug-selling accounts, with 83.5% ...

A qualitative study of three high-profile women shows that recovering from gendered online abuse is a strategic, ongoing process rather ...

A mixed-method corpus study of 500,599 Google Play reviews shows Indonesian swearing performs catharsis, complaint, humor, solidarity, intensification, and ideological ...

Researchers have developed a few-shot AI framework that combines agentic retrieval-augmented generation with multi-agent data augmentation to detect coded dog-whistle ...

A new study shows that fine-tuning large language models substantially outperforms prompting and classical baselines for detecting which group online ...

A new study of 100,000 tweets shows that LLaMA 3.2 refuses about 10 percent of paraphrasing requests, enforcing a narrow, ...

A new fairness-regularized framework called CoR-Hate retrieves real counterfactual examples from corpus data to reduce identity bias in hate-speech detection ...

A new study systematically tests early, late, and hybrid fusion strategies for combining image and text in AI systems that ...

Researchers have developed a reinforcement learning method that trains multimodal AI models to detect hateful and propagandistic memes in English ...