Automated README summary
Automatically summarized from the README on 29 Oct 2026; not a run by jevbooks. Run instructions and author-reported outcomes still need verification.
What it is
Open MedJev evaluates MedJev Flash 9B, an open System-One model, on medical decision benchmarks zero-shot and locally on Apple silicon, comparing against zero-shot LLMs, recalibration, and supervised baselines.
How it uses Jev
Each benchmark is cast as a Jev request: a state (text or JSON) plus typed questions, either a Choice over named options or a Noul (probability a statement is true). The model returns a full probability distribution from one forward pass without generating text; outputs are cached and scored with accuracy, AUROC/AUPRC, ECE, Brier and bootstrap CIs.
Primitives:choicenoul
Technique worth stealing
Casting every medical benchmark as a Jev request (state plus typed Choice/Noul questions) so one forward pass yields calibrated probabilities, cached by exact request.
Try it
src/scripts/run_all.sh, or LIMIT=500 src/scripts/run_all.sh for a quick run.
Model judgments from README excerpts
Project signals use repository metadata, an existing summary and the first 700 README characters, not the full project. “Not found” means not detected in that material; it does not establish absence. These are the original judgments with recorded models and dates, not source audits or run results.
Models & evaluation dates
- Inclusion gate
jev-1.14.0- Project signals
jev-1.14.0- Category
jev-1.14.0- Pattern tags
jev-1.14.0
Evidence
These signals use repository metadata, an existing summary and at most 700 cleaned README characters. ≥ 0.60 means reported in that excerpt; ≤ 0.40 means not found there. The middle band is inconclusive. None is a source-code audit or a test result. The runnable-tool question excludes research experiments; the recommendation question asks for a learning shortlist, not a quality verdict.
- Jev is core to the projectFound in material0.95
- Shows a System One patternunclear0.48
- Describes handling uncertaintynot found in the excerpt0.02
- Reports measured numbersnot found in the excerpt0.02
- Is a runnable toolnot found in the excerpt0.13
- Top-10 learning shortlist signalunclear0.46
- Is a replica of the modelunclear0.57
- Solves a common problem (0–2)score on a 0–2 scale0.75 / 2
- About JevFound in material0.98
no-judgment band (0.40–0.60)
no-judgment band (0.40–0.60)
The Jev model and date for each judgment stage are listed above. Card written by DeepSeek V4.1 Flash from the README on 29 Oct 2026.