USR: An Unsupervised and Reference Free Evaluation Metric for Dialog Generation

Benchmark Model Rank Results
dialogue-evaluation-on-usr-topicalchatUSR#3Spearman Correlation: 0.4192Pearson Correlation: 0.4220
dialogue-evaluation-on-usr-topicalchatUSR - DR (x = c)#4Spearman Correlation: 0.3245Pearson Correlation: 0.4068
dialogue-evaluation-on-usr-topicalchatUSR - MLM#5Spearman Correlation: 0.3086Pearson Correlation: 0.3345
dialogue-evaluation-on-usr-topicalchatUSR - DR (x = f)#6Spearman Correlation: 0.1419Pearson Correlation: 0.3221