Currently, completion is evaluated against randomly cut-off postfixes, where|prefix| > 2, but otherwise unbound. In order to provide a more informative metric for completion evaluation that also takes the new multi-suggestion capabilities into account, it would be preferencial to measure the postfix quality for fixed prefix lengths of 1 through 5 characters, within the first three completion results.
Currently, completion is evaluated against randomly cut-off postfixes, where
|prefix| > 2
, but otherwise unbound. In order to provide a more informative metric for completion evaluation that also takes the new multi-suggestion capabilities into account, it would be preferencial to measure the postfix quality for fixed prefix lengths of 1 through 5 characters, within the first three completion results.