# medical-jev-4b decision model by JuneYao

> medical-jev-4b scores supplied medical answer choices for offline research. Read its option limits, uncalibrated results and training-data caveats.

Updated 2026-11-12 (Australia/Adelaide)

## Quick answers

### What is medical-jev-4b by JuneYao?

[medical-jev-4b](<https://decisionmodels.ai/models/medical-jev>): A Qwen3.5-4B LoRA for supplied medical answer choices, mainly in Chinese.

[medical-jev-4b facts](<https://decisionmodels.ai/models/medical-jev>)

[Source](<https://huggingface.co/JuneYao/medical-jev-4b>)

### Who makes medical-jev-4b?

[medical-jev-4b](<https://decisionmodels.ai/models/medical-jev>): JuneYao

[medical-jev-4b facts](<https://decisionmodels.ai/models/medical-jev>)

[Source](<https://huggingface.co/JuneYao/medical-jev-4b>)

By JuneYao · Weights upload: 2026-11-02 · OPEN

A Qwen3.5-4B LoRA for supplied medical answer choices, mainly in Chinese. It reads next-token option logits in one forward pass and returns probabilities without an explanation. This is independent of TypeSafe’s hosted [Jev](<https://decisionmodels.ai/models/jev>). The released adapter is for teaching and offline research, not clinical use or triage.

## Facts

- Inputs: Case text, instructions and supplied choices; mainly Chinese, with other languages represented
- Limits: Released runtime: 52 options and 8,192 tokens; larger menus require the repository’s experimental tournament mode
- Licence: MIT code and adapter; Apache-2.0 base, with separate upstream data terms
- Access: Public custom-format LoRA and inference bundle; not PEFT-compatible
- Provenance: Catch-up weights released 2 Nov; source repository published 7 Nov

## Claims and sources

- **VENDOR-REPORTED**: The maker reports 1,028 of 1,514 benchmark questions correct (67.90%) with the released 52-option runtime; 100 unsupported questions count as wrong. Experimental tournament handling raises this to 1,050 (69.35%). The benchmark was repeatedly used for development and is not a blind test. [source](<https://github.com/JuneYaooo/medical-jev-train>)
- **VENDOR-REPORTED**: The maker reports 40 of 100 triage questions correct and uncalibrated benchmark ECE around 0.13. The original training file is lost; the public pool was reconstructed and may include sources from a later 1,557-item evaluation. That evaluation is not an untouched test of the released adapter. [source](<https://github.com/JuneYaooo/medical-jev-train>)

## Links

- [Source: Model card and adapter](<https://huggingface.co/JuneYao/medical-jev-4b>)
- [Source: Source and evaluation](<https://github.com/JuneYaooo/medical-jev-train>)
- [Source: Released runtime record](<https://github.com/JuneYaooo/medical-jev-train/blob/main/docs/results/v1/package-validation.json>)

## Related projects

### Medical answer-choice research

[Medical answer-choice research](<https://decisionmodels.ai/projects#project-medical-jev>)

[Source: JuneYaooo/medical-jev-train](<https://github.com/JuneYaooo/medical-jev-train>)

Training, inference and evaluation code for the [medical-jev-4b](<https://decisionmodels.ai/models/medical-jev>) adapter. The published GPU package check records 1,028 correct answers out of 1,514, with 100 over-limit questions counted wrong. Experimental tournament handling produces a different result, 1,050 correct.

- Creator: JuneYao

- Stars: 0

- Licence: MIT

- Model: [medical-jev-4b](<https://decisionmodels.ai/models/medical-jev>)

- Tags: medical education, fine-tuning

Author-run records, not a fresh reproduction. The benchmark was reused during development; the original training file is lost and the public training pool is reconstructed. The model is uncalibrated and is not for clinical use or triage.

- [Source: GitHub](<https://github.com/JuneYaooo/medical-jev-train>)
- [Source: Licence](<https://github.com/JuneYaooo/medical-jev-train/blob/main/LICENSE>)
- [Source: Inference code](<https://github.com/JuneYaooo/medical-jev-train/blob/main/src/jevtrain/core/inference.py>)
- [Source: GPU package check](<https://github.com/JuneYaooo/medical-jev-train/blob/main/docs/results/v1/package-validation.json>)
- [Source: Evaluation and limits](<https://github.com/JuneYaooo/medical-jev-train/blob/main/docs/EVALUATION.md>)