# The problem with LLMs as judges — retained record

> Retained reddit record from reddit:AI_Agents.

Canonical: https://highsignal.app/data/records/0d2a262d72756eb6

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:AI_Agents)  published 2026-09-29 16:16 ingested 2026-09-30 02:37

[original source ↗](https://www.reddit.com/r/AI_Agents/comments/1wtegk9/the_problem_with_llms_as_judges/)

## retained text



```
I've been playing with Jev, and it solves a big problem I've been noticing for a long time: using LLMs as a judge isn't very efficient.

LLMs rationalize everything. They're also eager to please, so they'll sometimes make things up just to meet the expected standard.

Jev's approach to structured data is pretty interesting. It's still probabilistic, but getting a mathematical result is much better than getting a diarrhea of words explaining a decision.

This feels like a step in the right direction.

Now the question is, will we see open source alternatives to Jev?
```



document fetched 2026-09-30 00:19 · canonical https://reddit.com/r/AI_Agents/comments/1wtegk9/the_problem_with_llms_as_judges
