Can Jev judge passage relevance?

An evaluation of Jev’s Choice, Score, and Noul primitives on the LLMJudge benchmark, with human labels and published LLM judges as reference points.

Study limitation

This is not a blind challenge submission. Test labels and benchmark results were public before this experiment. We expect Jev to have seen the public LLMJudge results during training, although its training data is unverified. These scores may reflect benchmark exposure.

What the labels say

Native Choice comes closest to the paper’s best four-grade judge. At the primary answer-bearing boundary, it reaches κ 0.4180 against a published best of 0.4280. Development-tuned Score and Noul thresholds both reduce test agreement.

Four-grade agreement

Cohen’s κ against human grades 0–3

  1. Choice0.2617
  2. Score0.2249
  3. Noul0.2589
  4. Paper best0.2863

Answer-bearing decision

Cohen’s κ for grades 0–1 versus 2–3

  1. Choice0.4180
  2. Score0.4059
  3. Noul0.3981
  4. Paper best0.4280

Each paper-best bar is the strongest of the paper’s 42 submitted judges for that metric; the winning submission differs by column. Bars start at zero and share a 0.5 κ scale.

All judgment boundaries

The 0–1 versus 2–3 split is the primary test of whether a passage provides an answer.

JudgeFour grades0 | 1–30–1 | 2–30–2 | 3
Jev ScoreNative0.22490.33670.40590.3099
Jev ScoreCalibrated0.21180.35920.36560.2907
Jev NoulNative0.25890.41490.39810.3103
Jev NoulCalibrated0.24510.38470.34020.3103
Paper best per column0.28630.42280.42800.3215

Paper winners: willia-umbrela1 (four grades), Olz-gpt4o (0 | 1–3), h2oloo-fewself (0–1 | 2–3), and willia-umbrela3 (0–2 | 3). Jev Choice exceeds the paper’s best only on the 0–2 | 3 boundary: 0.3308 versus 0.3215.

Do the labels preserve system order?

We scored the same 35 passage retrieval runs with human and Jev labels using trec_eval nDCG@10, then compared their rankings.

Read these side by side, not as a leaderboard. Our calculation does not reproduce the paper’s published τ and ρ for released submissions. The Jev values are independently computed under the convention described in the report.
JudgeKendall τSpearman ρ
Jev ChoiceNative0.92840.9891
Jev ScoreNative0.93510.9915
Jev ScoreCalibrated0.90240.9835
Jev NoulNative0.94280.9922
Jev NoulCalibrated0.92260.9874
Paper best per column0.95160.9919

Paper winners: prophet-setting2 for τ and TREMA-4prompts for ρ.

How the experiment works

The benchmark provides graded human judgments for 25 development and 25 test queries. The three primitives saw identical query and passage text, with frozen question wording.

Five questions, one request

  • ChoiceSelects one of the four benchmark grade definitions.
  • ScorePlaces the passage on an ordered four-level scale.
  • Noul × 3Tests relatedness, a partial answer, and a dedicated exact answer independently.

Native and calibrated labels

Native Score uses cutoffs 0.5 / 1.5 / 2.5; native Noul uses 0.5 on each answer. Calibration searched 0.05 steps to maximize four-grade κ on development data only. The selected Score cutoffs were 1.4 / 2.1 / 2.25 and Noul cutoffs were 0.8 / 0.9 / 0.5. They were applied once to test.

Evaluation

Four-grade and binary κ, nominal α, confusion matrices, grade distributions, and 95% query-cluster bootstrap intervals.

Coverage

All 4,423 test pairs have Jev responses. The report also compares 33 released LLM judges and 35 passage runs.

Limitations

Jev saw shortened head-and-tail text for 44 overlength passages. Test labels are public, and α conventions differ from the paper.

Read the full record

The site summarizes one experiment. The complete numeric report, raw Jev answers, threshold configuration, and source hashes are available for inspection.