BERTScore is Unfair: On Social Bias in Language Model-Based Metrics for Text Generation.

Tianxiang Sun, Junliang He, Xipeng Qiu, Xuanjing Huang

Browse the full EMNLP paper archive.