# Clinical Intent Alignment on HealthBench

Source: [Clinical Intent Alignment on HealthBench](<https://www.sota2.com/research/sota/clinical-intent-alignment-on-healthbench>)

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

Data updated: 2026-02-26T05:38:21.697Z

Evaluation of model performance on the Clinical Intent Alignment task using the HealthBench dataset, typically reporting CIA percentage and associated statistical significance.

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 | Clinical Intent Alignment |
| Dataset (including any stated split) | HealthBench |
| Metric | CIA |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 3 of 3 |
| 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

- [CIA](<https://www.sota2.com/research/sota/clinical-intent-alignment-on-healthbench.md?metric=CIA>): Higher is better; unit: PERCENT.

- [95% CI](<https://www.sota2.com/research/sota/clinical-intent-alignment-on-healthbench.md?metric=95%25%20CI>): Higher is better; unit: OTHER.

- [p-value](<https://www.sota2.com/research/sota/clinical-intent-alignment-on-healthbench.md?metric=p-value>): Lower is better; unit: SCALAR.

## Recorded results

| Method | CIA | Result date | Report paper |
| --- | --- | --- | --- |
| retrieval-augmented multi-agent framework | 60.12 | Not recorded | 6971af37e25ad2fa9d1edbc9 |
| GPT-4o Rubric | 55.16 | Not recorded | 6971af37e25ad2fa9d1edbc9 |
| Generic Rubric | 19.66 | Not recorded | 6971af37e25ad2fa9d1edbc9 |

### retrieval-augmented multi-agent framework

CIA: 60.12. Result ID: 6996763e5698665ec4bdbcd4.

[Reporting paper](<https://www.sota2.com/research/paper/6971af37e25ad2fa9d1edbc9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6971af37e25ad2fa9d1edbc9.md>)

Recorded evaluation configuration: {"Rubric":"Ours"}. This record does not assert independent reproduction.

### GPT-4o Rubric

CIA: 55.16. Result ID: 6996763e5698665ec4bdbcd3.

[Reporting paper](<https://www.sota2.com/research/paper/6971af37e25ad2fa9d1edbc9.md>)

Recorded evaluation configuration: {"Rubric":"GPT-4o"}. This record does not assert independent reproduction.

### Generic Rubric

CIA: 19.66. Result ID: 6996763e5698665ec4bdbcd2.

[Reporting paper](<https://www.sota2.com/research/paper/6971af37e25ad2fa9d1edbc9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72791e209131cea5696b.md>)

Recorded evaluation configuration: {"Rubric":"Generic"}. This record does not assert independent reproduction.

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