What it does
The feature approach consumed 34.1 million input tokens for $1.43 and produced 25 times as many false positives, clarifying when decomposition is worthwhile.
Articles & threads
One judge call vs twelve dimension scores
Agent Journal compares one holistic judge call with twelve Jev-derived features across three classification tasks.
Bookmark: One judge call vs twelve dimension scores
Keep this in your collection.
Leave a noteWhat would you try with this?
: One judge call vs twelve dimension scores
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
How you can use it
When building an app to sort written replies, test one clear AI question first. Use replies whose correct answers you already know. This article shows why asking twelve smaller questions is not always better: it needs a developer to write code that learns how to weigh the scores, costs more, and gave more false alarms in some tests.