Matt Pocock @mattpocockuk

LLM-as-a-judge in a few lines of code.

Important points:

- LLM's are bad at numeric scales, so we ask it to return a text enum which we then convert to a scale.
- Super useful for situations where deterministic tests just can't work.

Feb 7, 2025 View on X →

https://x.com/mattpocockuk/status/1887943661419130974