Jon Bratseth

Is this repeatable? Is coverage 100% in both cases? Could you try with top-k-probability set to 1.0?

This works just fine: ``` 1.0 ```

Workaround until this is done: Put everything in one group by grouping on a constant.

Still too early.

What in the query is there 124 of? Do you have the stacktrace? If not: Do you have custom searchers?

You're right and we should have kept the track trace in this case. I've done that in https://github.com/vespa-engine/vespa/pull/29782, should be available by Friday. The other way to figure out the...

Yes, you're right. We have not prioritized this because the query side is so simple, but key/cert handling complicates things a bit so this might be worth while now.

Yes, this is a new feature, but one that makes sense. How about creating a separate rank feature ("matchAccuracy"?) that gives the term-weighted average of the closeness of the match...

I think this a suggestion to add a to_array which does the splitting?

Yes, thanks!