Towards Understanding and Mitigating Social Biases in Language Models

          Paul Pu Liang 1 Chiyu Wu 1 Louis-Philippe Morency 1 Ruslan Salakhutdinov 1

             Abstract                Zhao et al., 2017). More recently, language models (LMs)
    Warning: this paper contains model outputs that      are increasingly used in real-world applications such as text
    may be offensive or upsetting.               generation (Radford et al., 2019), dialog systems (Zhang
                         ...