# URL Health and Self-Correction on ExpertQA

Source: [URL Health and Self-Correction on ExpertQA](<https://www.sota2.com/research/sota/url-health-and-self-correction-on-expertqa>)

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

Data updated: 2026-04-06T08:59:55.367Z

Evaluation of model performance on the ExpertQA dataset focusing on URL health assessment (live/dead status) and the efficacy of self-correction mechanisms, measured by correction rounds and URL success rates.

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 | URL Health and Self-Correction |
| Dataset (including any stated split) | ExpertQA |
| Metric | Total URLs |
| Unit | SCALAR |
| 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

- [Total URLs](<https://www.sota2.com/research/sota/url-health-and-self-correction-on-expertqa.md?metric=Total%20URLs>): Higher is better; unit: SCALAR.

- [Avg URLs per Question](<https://www.sota2.com/research/sota/url-health-and-self-correction-on-expertqa.md?metric=Avg%20URLs%20per%20Question>): Higher is better; unit: SCALAR.

- [Live URL Rate](<https://www.sota2.com/research/sota/url-health-and-self-correction-on-expertqa.md?metric=Live%20URL%20Rate>): Higher is better; unit: PERCENT.

- [Dead URL Rate](<https://www.sota2.com/research/sota/url-health-and-self-correction-on-expertqa.md?metric=Dead%20URL%20Rate>): Lower is better; unit: PERCENT.

- [Likely Hallucination Rate](<https://www.sota2.com/research/sota/url-health-and-self-correction-on-expertqa.md?metric=Likely%20Hallucination%20Rate>): Lower is better; unit: PERCENT.

- [Unknown URL Status Rate](<https://www.sota2.com/research/sota/url-health-and-self-correction-on-expertqa.md?metric=Unknown%20URL%20Status%20Rate>): Lower is better; unit: PERCENT.

- [Avg Live URLs per Question](<https://www.sota2.com/research/sota/url-health-and-self-correction-on-expertqa.md?metric=Avg%20Live%20URLs%20per%20Question>): Higher is better; unit: SCALAR.

- [Mean Correction Rounds](<https://www.sota2.com/research/sota/url-health-and-self-correction-on-expertqa.md?metric=Mean%20Correction%20Rounds>): Lower is better; unit: SCALAR.

- [Median Correction Rounds](<https://www.sota2.com/research/sota/url-health-and-self-correction-on-expertqa.md?metric=Median%20Correction%20Rounds>): Lower is better; unit: SCALAR.

- [Max Correction Rounds](<https://www.sota2.com/research/sota/url-health-and-self-correction-on-expertqa.md?metric=Max%20Correction%20Rounds>): Lower is better; unit: SCALAR.

## Recorded results

| Method | Total URLs | Result date | Report paper |
| --- | --- | --- | --- |
| Claude | 7985 | Not recorded | 69d33df87d0172140ec7367f |
| GPT-5.1 | 4829 | Not recorded | 69d33df87d0172140ec7367f |
| Gemini | 4203 | Not recorded | 69d33df87d0172140ec7367f |

### Claude

Total URLs: 7985. Result ID: 69d3760847f34ce924bb741c.

[Reporting paper](<https://www.sota2.com/research/paper/69d33df87d0172140ec7367f.md>)

Recorded evaluation configuration: {"Agentic tool usage":"urlhealth"}. This record does not assert independent reproduction.

### GPT-5.1

Total URLs: 4829. Result ID: 69d3760847f34ce924bb741e.

[Reporting paper](<https://www.sota2.com/research/paper/69d33df87d0172140ec7367f.md>)

Recorded evaluation configuration: {"Agentic tool usage":"urlhealth"}. This record does not assert independent reproduction.

### Gemini

Total URLs: 4203. Result ID: 69d3760847f34ce924bb741d.

[Reporting paper](<https://www.sota2.com/research/paper/69d33df87d0172140ec7367f.md>)

Recorded evaluation configuration: {"Agentic tool usage":"urlhealth"}. This record does not assert independent reproduction.

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