# jev-orderby-bench

Measures whether ORDER BY over a Jev probability is defensible: pairwise inversion, Score ordinality against a human grade, calibration, and wording invariants under a pre-registered gate; passes on 20 Newsgroups topics, fails four of six conditions on Amazon ESCI product relevance, and shows that a 40-row batched state through a DuckDB extension fails the ranking gate one row per request passes.

- URL: https://github.com/yodablocks/jev-orderby-bench
- Section: Benchmarks, evals and calibration
- Type: project
- Stars: 0
- Pushed: 2026-09-20T05:39:46Z
- Language: Python
- License: MIT
- Claim in the entry: pre-registered
- First seen: 2026-09-19
- Form: Dataset or benchmark
- Useful for: Benchmarks and evals (0.90), Output verification (0.89), Search reranking (0.76)

Page: https://awesomejev.vercel.app/p/jev-orderby-bench/