daltonj
daltonj
Does this mean there are new Anserini RM3 results corrected? And is this 'bug' actually a feature?
It doesn't seem right to not fix a bug because it would change numbers. Isn't this the correct, desired outcome of a bug fix? Fix the bug, update the tests..?...
Can we also update this issue quantifying the impact on MAP, and other standard metrics. How big is it? I expect the tie breaking to be small. But what about...
Thanks. I appreciate the fast turn around. Is there a corresponding pull request / diff to review the RM3 changes? Maybe I could take a stab at reviewing the RM3...