# medgemma-decide-4b-v2 by Amaljith

> medgemma-decide-4b-v2 scores medical text options for research only. Read its class-balance gains, regressions and HAI-DEF licence terms.

Updated 2026-11-12 (Australia/Adelaide)

## Quick answers

### What is medgemma-decide-4b-v2 by Amaljith?

[medgemma-decide-4b-v2](<https://decisionmodels.ai/models/medgemma-decide>): An English medical-text decision fine-tune continued from a MedGemma 1.5 4B derivative.

[medgemma-decide-4b-v2 facts](<https://decisionmodels.ai/models/medgemma-decide>)

[Source](<https://huggingface.co/Amaljith/medgemma-decide-4b-v2>)

### Who makes medgemma-decide-4b-v2?

[medgemma-decide-4b-v2](<https://decisionmodels.ai/models/medgemma-decide>): Amaljith

[medgemma-decide-4b-v2 facts](<https://decisionmodels.ai/models/medgemma-decide>)

[Source](<https://huggingface.co/Amaljith/medgemma-decide-4b-v2>)

### Are medgemma-decide-4b-v2's weights open?

[medgemma-decide-4b-v2](<https://decisionmodels.ai/models/medgemma-decide>): Public merged weights, joint decision head, adapter, inference code and per-question temperatures Licence: Health AI Developer Foundations terms; derivative notice included

[medgemma-decide-4b-v2 facts](<https://decisionmodels.ai/models/medgemma-decide>)

[Source](<https://huggingface.co/Amaljith/medgemma-decide-4b-v2>)

By Amaljith · Weights upload: 2026-11-07 · RESEARCH ONLY

An English medical-text decision fine-tune continued from a MedGemma 1.5 4B derivative. It scores typed options without generated text. Minority-class oversampling changes the decision balance, with gains and regressions across questions. Research use only, not for diagnosis, triage, clinical decisions or patient care.

## Facts

- Base: unsloth/medgemma-1.5-4b-it, continued from the maker’s first released checkpoint
- Inputs: English medical text and supplied typed options; text only
- Training: Fresh rank-16 LoRA; one epoch, 728 steps, 51.7 minutes on an A100 80 GB, according to the card
- Licence: Health AI Developer Foundations terms; derivative notice included
- Access: Public merged weights, joint decision head, adapter, inference code and per-question temperatures

## Claims and sources

- **VENDOR-REPORTED**: The maker reports medline/adverse balanced accuracy rising from 0.625 to 0.825, above its TF-IDF baseline floor of 0.721. Macro balanced accuracy across sources falls from 0.689 to 0.686. The evaluation has 15,320 records and 19,480 decisions; balanced accuracy uses a prior-adjusted decision rule. [source](<https://huggingface.co/Amaljith/medgemma-decide-4b-v2>)
- **VENDOR-REPORTED**: The card reports PubMedQA balanced accuracy falling from 0.578 to 0.488 under that rule. Public exam sets, MeSH labels and one seed per run limit interpretation; seed variance is unmeasured. This is not clinical validation. [source](<https://huggingface.co/Amaljith/medgemma-decide-4b-v2>)

## Links

- [Source: Model card and weights](<https://huggingface.co/Amaljith/medgemma-decide-4b-v2>)
- [Source: Per-question results](<https://huggingface.co/Amaljith/medgemma-decide-4b-v2/raw/main/results.json>)
- [Source: Derivative notice](<https://huggingface.co/Amaljith/medgemma-decide-4b-v2/blob/main/NOTICE>)

## Related projects

No related projects are listed in the registry.