# Medical Question Answering on MedQA, MedMCQA, PubMedQA, MMLU-Pro-med, and GPQA-med (test)

Source: [Medical Question Answering on MedQA, MedMCQA, PubMedQA, MMLU-Pro-med, and GPQA-med (test)](<https://www.sota2.com/research/sota/medical-question-answering-on-medqa-medmcqa-pubmedqa-mmlu-pro-med-and-gpqa-med-test>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-04-28T08:20:39.153Z

Evaluation of model performance on a composite set of medical question answering datasets, including MedQA, MedMCQA, PubMedQA, MMLU-Pro-med, and the held-out test split of GPQA-med.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Medical Question Answering |
| Dataset (including any stated split) | MedQA, MedMCQA, PubMedQA, MMLU-Pro-med, GPQA-med held-out (test) |
| Metric | Accuracy (MedQA) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 9 of 9 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Accuracy (MedQA)](<https://www.sota2.com/research/sota/medical-question-answering-on-medqa-medmcqa-pubmedqa-mmlu-pro-med-and-gpqa-med-test.md?metric=Accuracy%20(MedQA)>): Higher is better; unit: PERCENT.

- [Accuracy (MedMCQA)](<https://www.sota2.com/research/sota/medical-question-answering-on-medqa-medmcqa-pubmedqa-mmlu-pro-med-and-gpqa-med-test.md?metric=Accuracy%20(MedMCQA)>): Higher is better; unit: PERCENT.

- [Accuracy (PubMedQA)](<https://www.sota2.com/research/sota/medical-question-answering-on-medqa-medmcqa-pubmedqa-mmlu-pro-med-and-gpqa-med-test.md?metric=Accuracy%20(PubMedQA)>): Higher is better; unit: PERCENT.

- [Accuracy (MMLU-Pro-med)](<https://www.sota2.com/research/sota/medical-question-answering-on-medqa-medmcqa-pubmedqa-mmlu-pro-med-and-gpqa-med-test.md?metric=Accuracy%20(MMLU-Pro-med)>): Higher is better; unit: PERCENT.

- [Accuracy (GPQA-med)](<https://www.sota2.com/research/sota/medical-question-answering-on-medqa-medmcqa-pubmedqa-mmlu-pro-med-and-gpqa-med-test.md?metric=Accuracy%20(GPQA-med)>): Higher is better; unit: PERCENT.

- [Average Accuracy](<https://www.sota2.com/research/sota/medical-question-answering-on-medqa-medmcqa-pubmedqa-mmlu-pro-med-and-gpqa-med-test.md?metric=Average%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy (MedQA) | Result date | Report paper |
| --- | --- | --- | --- |
| For-Value | 57.61 | Not recorded | 68fa7aa51e209131cea845cf |
| Hessian-free | 57.02 | Not recorded | 68fa7aa51e209131cea845cf |
| HyperINF | 56.94 | Not recorded | 68fa7aa51e209131cea845cf |
| Base | 56.84 | Not recorded | 68fa7aa51e209131cea845cf |
| For-Value | 56.8 | Not recorded | 68fa7aa51e209131cea845cf |
| DataInf | 56.61 | Not recorded | 68fa7aa51e209131cea845cf |
| Hessian-free | 55.41 | Not recorded | 68fa7aa51e209131cea845cf |
| DataInf | 55.39 | Not recorded | 68fa7aa51e209131cea845cf |
| HyperINF | 55.15 | Not recorded | 68fa7aa51e209131cea845cf |

### For-Value

Accuracy (MedQA): 57.61. Result ID: 69f06d4947f34ce924d63ebd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7aa51e209131cea845cf.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7aa51e209131cea845cf.md>)

[Implementation](<https://github.com/vengdeng/For-Value-Efficient-Forward-Only-Data-Valuation-for-finetuning>)

Recorded evaluation configuration: {"Backbone":"Llama-3.1-8B-Ins","Data Selection Ratio":"10%","Time":"0.8 (h)"}. This record does not assert independent reproduction.

### Hessian-free

Accuracy (MedQA): 57.02. Result ID: 69f06d4947f34ce924d63eba.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7aa51e209131cea845cf.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa68ba1e209131cea1dca1.md>)

Recorded evaluation configuration: {"Backbone":"Llama-3.1-8B-Ins","Data Selection Ratio":"10%","Time":"2.0 (h)"}. This record does not assert independent reproduction.

### HyperINF

Accuracy (MedQA): 56.94. Result ID: 69f06d4947f34ce924d63ebb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7aa51e209131cea845cf.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79a11e209131cea7e8a7.md>)

Recorded evaluation configuration: {"Backbone":"Llama-3.1-8B-Ins","Data Selection Ratio":"10%","Time":"5.3 (h)"}. This record does not assert independent reproduction.

### Base

Accuracy (MedQA): 56.84. Result ID: 69f06d4847f34ce924d63eb5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7aa51e209131cea845cf.md>)

Recorded evaluation configuration: {"Backbone":"Llama-3.1-8B-Ins"}. This record does not assert independent reproduction.

### For-Value

Accuracy (MedQA): 56.8. Result ID: 69f06d4947f34ce924d63eb9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7aa51e209131cea845cf.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7aa51e209131cea845cf.md>)

[Implementation](<https://github.com/vengdeng/For-Value-Efficient-Forward-Only-Data-Valuation-for-finetuning>)

Recorded evaluation configuration: {"Backbone":"Llama-3.1-8B-Ins","Data Selection Ratio":"5%","Time":"0.8 (h)"}. This record does not assert independent reproduction.

### DataInf

Accuracy (MedQA): 56.61. Result ID: 69f06d4947f34ce924d63ebc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7aa51e209131cea845cf.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72821e209131cea56e03.md>)

Recorded evaluation configuration: {"Backbone":"Llama-3.1-8B-Ins","Data Selection Ratio":"10%","Time":"4.1 (h)"}. This record does not assert independent reproduction.

### Hessian-free

Accuracy (MedQA): 55.41. Result ID: 69f06d4947f34ce924d63eb6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7aa51e209131cea845cf.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa68ba1e209131cea1dca1.md>)

Recorded evaluation configuration: {"Backbone":"Llama-3.1-8B-Ins","Data Selection Ratio":"5%","Time":"2.0 (h)"}. This record does not assert independent reproduction.

### DataInf

Accuracy (MedQA): 55.39. Result ID: 69f06d4947f34ce924d63eb8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7aa51e209131cea845cf.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72821e209131cea56e03.md>)

Recorded evaluation configuration: {"Backbone":"Llama-3.1-8B-Ins","Data Selection Ratio":"5%","Time":"4.1 (h)"}. This record does not assert independent reproduction.

### HyperINF

Accuracy (MedQA): 55.15. Result ID: 69f06d4947f34ce924d63eb7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7aa51e209131cea845cf.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79a11e209131cea7e8a7.md>)

Recorded evaluation configuration: {"Backbone":"Llama-3.1-8B-Ins","Data Selection Ratio":"5%","Time":"5.3 (h)"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.