Evaluation Run 20260923T023046
Rubric 2026-09-23b · typesafe/jev-1.13 · 431 of 431 answered · 16.1s · $0.0417
No entry file was written. Results live in this run's ledger until Phase 4 adds the write path, after calibration sets the thresholds.
Per-dimension distribution
| Dimension | Type | Distribution | Mean confidence | Agrees |
|---|---|---|---|---|
answer_key_correct | noul | mean 0.94 · level 1 holds 429/431 FLAT | — | — |
difficulty | score | mean 1.22 · level 1 holds 243/431 | 0.42 | 304/431 |
distractor_pull | score | mean 1.41 · level 1 holds 222/431 | 0.50 | — |
engagement | score | mean 1.36 · level 2 holds 213/431 | 0.63 | — |
one_defensible_answer | noul | mean 0.84 · level 1 holds 296/302 FLAT | — | — |
options_scan_clean | score | mean 1.58 · level 2 holds 303/431 | 0.55 | — |
passage_realism | score | mean 1.35 · level 1 holds 177/302 | 0.56 | — |
reads_in_one_pass | score | mean 1.85 · level 2 holds 419/431 FLAT | 0.78 | — |
sat_choice_conventions | score | mean 1.83 · level 2 holds 411/431 FLAT | 0.77 | — |
sat_stem_conventions | score | mean 1.83 · level 2 holds 406/431 FLAT | 0.85 | — |
section | choice | reading 178/431 | 1.00 | 431/431 |
single_move | noul | mean 0.74 · level 1 holds 365/431 FLAT | — | — |
skill_code | choice | info-and-ideas 133/431 | 0.95 | 406/429 |
stem_self_contained | noul | mean 0.95 · level 1 holds 431/431 FLAT | — | — |
survives_speed | noul | mean 0.70 · level 1 holds 418/431 FLAT | — | — |
tested_construct_is_sat | noul | mean 0.96 · level 1 holds 431/431 FLAT | — | — |
A dimension marked FLAT holds more than 80% of the run on one level. That is either a clean bank or a dimension reporting a constant, and only a negative control tells them apart.
Run
| Served by | {'typesafe/jev-1.13-20260917 via TypeSafe': 431} |
| Input tokens | 991,932 |
| Failed | 0 |
| Preflight notes | 0 |
| Skipped entries | 0 |