一作ACL 2026Poster· Natural Language Processing
Reinforcement Learning–Guided Adaptive Tuning for Out-of-Distribution Harmful Text Detection
Mengyu Xiang · Tinghao Chen · Boxu Han · Qiudan Li · shu wu · Daniel Dajun Zeng
摘要
As social media grows, harmful information spreads rapidly across platforms and evolves over time, showing cross-platform and cross-temporal variations. Existing methods rely on fixed model parameters during training, which fail to handle substantial semantic discrepancies, leading to Out-Of-Dist...