2,590 followers
@jobergum @macavaney @ntonellotto Correlations between ranking phases... We did a large scale study in the LTR era - https://t.co/Prs4GAA3AD. One observation was that learned models did not generalise across different candidate set sizes. A simple example: