Researchers have developed a novel dual-threshold hard example mining strategy to improve the performance of offensive comment detection models across different Chinese social media platforms. The proposed method involves finetuning a RoBERTa model and then further adapting it using a small set of manually labeled hard examples, identified from unlabeled corpora based on prediction confidence. This approach effectively addresses the performance degradation typically seen when models are deployed across diverse platforms like Weibo, Xiaohongshu, Tieba, and Zhihu, demonstrating significant performance gains. AI
IMPACT Improves cross-platform AI model adaptation for content moderation tasks.
RANK_REASON The cluster contains a research paper detailing a new method for AI model adaptation.
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
