Fine-tuned Qwen-0.5B slightly edges out DistilBERT and RoBERTa on a balanced hate speech corpus, and few-shot prompting substantially improves Gemma-3-1B over zero-shot.
Now you see me: Vision-enhanced bert for obfuscated text abuse detection,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
From BERT to Qwen: Hate Detection across architectures
Fine-tuned Qwen-0.5B slightly edges out DistilBERT and RoBERTa on a balanced hate speech corpus, and few-shot prompting substantially improves Gemma-3-1B over zero-shot.