Automated README summary
Automatically summarized from the README on 30 Sept 2026; not a run by jevbooks. Run instructions and author-reported outcomes still need verification.
What it is
An independent, pre-registered evaluation of TypeSafe's Jev model on SciFact scientific claim verification: 300 expert-labelled biomedical claims checked against 5,183 PubMed abstracts, with results, caveats and reproduction scripts.
How it uses Jev
Jev reranks BM25 abstracts via a Score question rating directness; for the top 3, a Choice question gives a supports/contradicts verdict and one Noul per sentence flags evidence. Thresholds are tuned on train only and applied in stage 04.
Primitives:scorechoicenoul
Technique worth stealing
Freeze question wording and thresholds on a training split before running the held-out dev split once.
Try it
uv sync; uv run python scripts/fetch_scifact.py; run scripts 01–05 with --split dev
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-1.13.0- Project signals
jev-1.13.0- Category
jev-1.13.0- Pattern tags
jev-1.13.0
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 material0.74
- Shows a System One patternnot found in the excerpt0.08
- Describes handling uncertaintynot found in the excerpt0.09
- Reports measured numbersFound in material0.96
- Is a runnable toolnot found in the excerpt0.04
- Top-10 learning shortlist signalnot found in the excerpt0.36
- Is a replica of the modelnot found in the excerpt0.04
- Solves a common problem (0–2)score on a 0–2 scale0.93 / 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 30 Sept 2026.