User repository on GitHub
Code for "OmniMed-Jev: Calibrating LVLM Confidence for Trustworthy Medical Multimodal Decisions via System One"
- Forks
- 0
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Models & Reimplementations
OmniMed-Jev: Calibrating LVLM Confidence for Trustworthy Medical Multimodal Decisions via System One (arXiv 2610.00381)
A decision-native interface that represents each medical decision as a Choice, Noul, or Score over a runtime candidate set and returns a calibrated distribution over it. Against a matched generative baseline it cuts calibration error by an order of magnitude, with counting the one task where the baseline leads.
- Stars
- 1
- Language
- Python
- License
- APACHE-2.0
- Format
- Article
- Published by
- Community
- Last verified
- Oct 2026
- Languages
- Python
Highlights
- Represents each decision as Choice, Noul, or Score over a runtime candidate set and returns a full distribution.
- Accepts diverse imaging modalities and covers different prediction tasks through one probability model.
- Reduces calibration error by up to an order of magnitude against a matched generative baseline.
- Reliability error falls by up to two times while point-prediction performance stays comparable.
- Counting is the one family where the generative baseline stays ahead.
Watch out
arXiv preprint posted September 30, 2026 (cs.LG) by Luyao Tang and Cheng Chen. The interface-controlled comparison cannot separate the decision interface from associated training differences, and the authors state the results are not evidence of clinical readiness.
Questions about this entry
Short answers drawn from the linked project or page.
It represents each medical decision as a Choice, Noul, or Score decision over a runtime-supplied candidate set and returns a full distribution over that set: mutually exclusive classes, binary presence of a finding, or a bounded ordered value.
It accepts diverse imaging modalities, covers different prediction tasks, and expresses them through one candidate-conditioned probability model, so heterogeneous outputs become comparable probabilities instead of task-specific strings.
In an interface-controlled comparison against a generative baseline trained on the same backbone, data, and schedule, its reported probabilities track observed correctness far more closely.
Calibration error falls by up to an order of magnitude and reliability error by up to two, while point-prediction performance remains comparable.
Counting is the one family where the generative baseline stays ahead, the authors report.
No. The paper states the results are not evidence of clinical readiness, and the comparison cannot separate the interface from associated training differences.
Code is available at github.com/lytang63/OmniMed-Jev.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
Hacker News: A one-token logprob wrapper that also classifies webcam frames.
allanrbo151 pointsA single function Jev-like wrapper for LLMs, including vision modelsRead the thread on Hacker NewsRelated terms
Glossary definitions related to this entry.
More Jev resources like this
Introduces Jeva, a decision-native architecture that encodes the request and each option separately and scores options symmetrically. It targets null-awareness, permutation invariance, option scalability, and cacheability, reporting near-perfect null detection and no decision flips under tested permutations.
Jev Models & ReimplementationsArticle#community#research#arxiv
Community
A System One model for Chinese-language decisions trained on a 10-million-example corpus and fine-tuned for medicine, legal, and finance. Pre-training exceeds closed-source Jev by 1.24% on general tasks with a 20.3x speedup; domain fine-tuning beats Jev in medicine by 4.0%.
Jev Models & ReimplementationsArticle#community#research#arxiv
Community
Presents PixelJev, a native-image decision interface that maps an image, a task instruction, and a runtime candidate set to a structured choice and candidate-conditioned probabilities. 64-shot source adaptation raises Pets accuracy from 60.13% to 92.40% across optimization seeds.
Jev Models & ReimplementationsArticle#community#research#arxiv
Community
1.3kGitHub stars
Reads a typed decision from one prefill of a pretrained instruction-tuned model by restricting the next-token distribution to option tokens, then corrects label and position bias with no gradient steps. Averaging over cyclic rotations lowers order-flip rates and raises accuracy on 11 of 11 models.
Jev Models & ReimplementationsArticle#community#research#arxiv
Community
nokia-applied-research/AnyJevPython
nokia-applied-research/AnyJevPythonSBERT2S1 converts biomedical Sentence-Transformers encoders into typed decision models with bi-encoder cross-head and prior-fused residual variants, alongside BIODECIDE and MEDLINE-S1. Retrieval training helps the prior-fused head in 10 of 15 comparisons after fine-tuning.
Jev Models & ReimplementationsArticle#community#research#arxiv
Community
LAVOIR extends the Laya single-pass decision encoder so one forward pass returns both the decision distribution and, for every candidate missing slot, the expected gain from asking the user about it. Its question policy matches a greedy oracle on seen schemas and is 14.1 points more accurate than never asking.
Jev Models & ReimplementationsArticle#community#research#arxiv
Community
moganai/lavoirPython
moganai/lavoirPythonFrom the Jev guides
Original write-ups on what this entry is for.
More than eighty preprints now evaluate, attack, and reproduce the typed-decision class. The broadest board, the coherence gap, the rejection bottleneck, and the attacks that still work.
Jev Practices & Patterns#research#evaluation#calibration
Read article
TypeSafe publishes evals, but the numbers that matter are the ones measured on your task. A tour of the independent benchmarks, field reports, and the recurring finding that question design dominates.
Jev Practices & Patterns#benchmarks#evaluation#calibration
Read article
Training recipes, looped architectures, and language ports: what the open side of the typed-decision class shipped this month, and how to pick by runtime and calibration instead of the leaderboard.
Jev Models & Reimplementations#open-models#training#local
Read article
From the Jev community
Posts from builders shipping with Jev right now.