Jev Judging
3 curated entries tagged judging.
All 3 Jev Judging
Filter these entries by what they are called or what they do.
No entries match those filters.
None of the entries shown here match. 0 more carry this tag — browse all 3 in the directory.
- A bias audit of typed decision models used as judges, running the CALM benchmark's bias tests on eight TDMs and seven LLM judges from 2024 and 2026. Every TDM is biased in three to eight of twelve tests, and emotional tone drops a typical TDM's accuracy from 99% to 42%.Jev Benchmarks & EvaluationsArticle#community#research#arxivCommunity
- A controlled evaluation of Jev as a source-support verifier for GPT-4.1-mini calculation traces on financial reports. Holding operands and arithmetic fixed while swapping citations between same-number cells reveals wrong-role citations that pass and valid alternative citations that are withheld.Jev Benchmarks & EvaluationsArticle#community#research#arxivCommunity
- An evaluation of Jev as a low-cost judge of agreement with physician-written radiology reports. A single-question configuration reaches Kendall correlations of 0.573 on RadEvalX and 0.398 on RadEvalExpert, and one support question per statement keeps expert agreement while using 43-45% fewer input tokens.Jev Benchmarks & EvaluationsArticle#community#research#arxivCommunity