RV-HATE: A New Framework for Implicit Hate Speech Detection Using Reinforcement Learning

A new detection framework named RV-HATE has been developed by researchers to tackle the issue of implicit hate speech by tailoring itself to the specific traits of various datasets. The evolution of hate speech, influenced by the internet and the anonymity it provides, makes detection increasingly difficult. Different datasets arise from various sources and platforms, each characterized by its own linguistic styles and social contexts. Previous research often relied on rigid methodologies that overlooked these unique data characteristics. RV-HATE features several specialized modules, each targeting specific linguistic or contextual aspects of hate speech. Utilizing reinforcement learning, the framework adjusts the weights that define each module’s impact, enhancing detection adaptability. This study has been published on arXiv with the identifier 2510.10971.