Automated README summary
Automatically summarized from the README on 25 Sept 2026; not a run by jevbooks. Run instructions and author-reported outcomes still need verification.
What it is
A benchmark harness that adds TypeSafe's Jev 1.13 and four fast LLMs to Stanford MedHELM's MedHallu results, testing whether a proposed answer misrepresents its PubMed abstract. For researchers comparing hallucination checks.
How it uses Jev
Jev answers one Noul question: 'Would the authors of the study described in knowledge say that answer misrepresents their findings?' on the abstract, question and answer. Its probability is thresholded at 0.65 to decide hallucination; the result is scored for accuracy and used to route confident items away from LLMs.
Primitives:noul
Technique worth stealing
Pre-register every test and follow-up check, then let Jev settle high-confidence items to cut LLM calls.
Try it
cd bench && uv run datasets/medhallu/prepare.py; uv run score.py runs/medhallu/test-v2 --threshold 0.65 --ref jev-1.13 --xlsx
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 material0.75
- Shows a System One patternFound in material0.80
- Describes handling uncertaintyFound in material0.66
- Reports measured numbersFound in material0.98
- Is a runnable toolnot found in the excerpt0.07
- Top-10 learning shortlist signalunclear0.49
- Is a replica of the modelnot found in the excerpt0.03
- Solves a common problem (0–2)score on a 0–2 scale1.39 / 2
- About JevFound in material0.99
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 25 Sept 2026.