# Vulnerability Generation on Vulnerability Generation Prompt Dataset (test)

Source: [Vulnerability Generation on Vulnerability Generation Prompt Dataset (test)](<https://www.sota2.com/research/sota/vulnerability-generation-on-vulnerability-generation-prompt-dataset-test>)

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

Data updated: 2026-06-11T08:50:01.678Z

Evaluation of model performance on generating vulnerabilities using the test split of the Vulnerability Generation Prompt Dataset, measured by the VGR metric.

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 | Vulnerability Generation |
| Dataset (including any stated split) | Vulnerability Generation Prompt Dataset (test) |
| Metric | VGR |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 15 of 15 |
| 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

- [VGR](<https://www.sota2.com/research/sota/vulnerability-generation-on-vulnerability-generation-prompt-dataset-test.md?metric=VGR>): Higher is better; unit: PERCENT.

## Recorded results

| Method | VGR | Result date | Report paper |
| --- | --- | --- | --- |
| GPT-3.5-Turbo | 68.7 | Not recorded | 6a28f38290966ade919c1a74 |
| GPT-3.5-Turbo | 52 | Not recorded | 6a28f38290966ade919c1a74 |
| GPT-4 | 40 | Not recorded | 6a28f38290966ade919c1a74 |
| Mean (All Models) | 37.4 | Not recorded | 6a28f38290966ade919c1a74 |
| GPT-4 | 30 | Not recorded | 6a28f38290966ade919c1a74 |
| Mean (All Models) | 27.5 | Not recorded | 6a28f38290966ade919c1a74 |
| CodeT5+ | 22.7 | Not recorded | 6a28f38290966ade919c1a74 |
| CodeLlama | 18.3 | Not recorded | 6a28f38290966ade919c1a74 |
| CodeT5+ | 15 | Not recorded | 6a28f38290966ade919c1a74 |
| CodeLlama | 13.1 | Not recorded | 6a28f38290966ade919c1a74 |
| CodeLlama | 12 | Not recorded | 6a28f38290966ade919c1a74 |
| Mean (All Models) | 3.5 | Not recorded | 6a28f38290966ade919c1a74 |
| CodeT5+ | 2 | Not recorded | 6a28f38290966ade919c1a74 |
| GPT-3.5-Turbo | 0 | Not recorded | 6a28f38290966ade919c1a74 |
| GPT-4 | 0 | Not recorded | 6a28f38290966ade919c1a74 |

### GPT-3.5-Turbo

VGR: 68.7. Result ID: 6a2a76b8fbc373592c27af89.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f38290966ade919c1a74.md>)

Recorded evaluation configuration: {"n (Sample Size)":"500","Prompt Condition":"Adversarial"}. This record does not assert independent reproduction.

### GPT-3.5-Turbo

VGR: 52. Result ID: 6a2a76b8fbc373592c27af8a.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f38290966ade919c1a74.md>)

Recorded evaluation configuration: {"n (Sample Size)":"500","Prompt Condition":"Mean"}. This record does not assert independent reproduction.

### GPT-4

VGR: 40. Result ID: 6a2a76b8fbc373592c27af8c.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f38290966ade919c1a74.md>)

Recorded evaluation configuration: {"n (Sample Size)":"300","Prompt Condition":"Adversarial"}. This record does not assert independent reproduction.

### Mean (All Models)

VGR: 37.4. Result ID: 6a2a76b8fbc373592c27af8f.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f38290966ade919c1a74.md>)

Recorded evaluation configuration: {"n (Sample Size)":"2,800","Prompt Condition":"Adversarial"}. This record does not assert independent reproduction.

### GPT-4

VGR: 30. Result ID: 6a2a76b8fbc373592c27af8d.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f38290966ade919c1a74.md>)

Recorded evaluation configuration: {"n (Sample Size)":"300","Prompt Condition":"Mean"}. This record does not assert independent reproduction.

### Mean (All Models)

VGR: 27.5. Result ID: 6a2a76b8fbc373592c27af90.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f38290966ade919c1a74.md>)

Recorded evaluation configuration: {"n (Sample Size)":"2,800","Prompt Condition":"Mean"}. This record does not assert independent reproduction.

### CodeT5+

VGR: 22.7. Result ID: 6a2a76b8fbc373592c27af83.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f38290966ade919c1a74.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f731e209131cea45aae.md>)

[Implementation](<https://github.com/salesforce/CodeT5/tree/main/CodeT5+>)

Recorded evaluation configuration: {"n (Sample Size)":"1,000","Prompt Condition":"Adversarial"}. This record does not assert independent reproduction.

### CodeLlama

VGR: 18.3. Result ID: 6a2a76b8fbc373592c27af86.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f38290966ade919c1a74.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71ff1e209131cea53ff7.md>)

[Implementation](<https://github.com/facebookresearch/codellama>)

Recorded evaluation configuration: {"n (Sample Size)":"1,000","Prompt Condition":"Adversarial"}. This record does not assert independent reproduction.

### CodeT5+

VGR: 15. Result ID: 6a2a76b8fbc373592c27af84.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f38290966ade919c1a74.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f731e209131cea45aae.md>)

[Implementation](<https://github.com/salesforce/CodeT5/tree/main/CodeT5+>)

Recorded evaluation configuration: {"n (Sample Size)":"1,000","Prompt Condition":"Mean"}. This record does not assert independent reproduction.

### CodeLlama

VGR: 13.1. Result ID: 6a2a76b8fbc373592c27af87.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f38290966ade919c1a74.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71ff1e209131cea53ff7.md>)

[Implementation](<https://github.com/facebookresearch/codellama>)

Recorded evaluation configuration: {"n (Sample Size)":"1,000","Prompt Condition":"Mean"}. This record does not assert independent reproduction.

### CodeLlama

VGR: 12. Result ID: 6a2a76b8fbc373592c27af85.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f38290966ade919c1a74.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71ff1e209131cea53ff7.md>)

[Implementation](<https://github.com/facebookresearch/codellama>)

Recorded evaluation configuration: {"n (Sample Size)":"1,000","Prompt Condition":"Baseline"}. This record does not assert independent reproduction.

### Mean (All Models)

VGR: 3.5. Result ID: 6a2a76b8fbc373592c27af8e.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f38290966ade919c1a74.md>)

Recorded evaluation configuration: {"n (Sample Size)":"2,800","Prompt Condition":"Baseline"}. This record does not assert independent reproduction.

### CodeT5+

VGR: 2. Result ID: 6a2a76b8fbc373592c27af82.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f38290966ade919c1a74.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f731e209131cea45aae.md>)

[Implementation](<https://github.com/salesforce/CodeT5/tree/main/CodeT5+>)

Recorded evaluation configuration: {"n (Sample Size)":"1,000","Prompt Condition":"Baseline"}. This record does not assert independent reproduction.

### GPT-3.5-Turbo

VGR: 0. Result ID: 6a2a76b8fbc373592c27af88.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f38290966ade919c1a74.md>)

Recorded evaluation configuration: {"n (Sample Size)":"500","Prompt Condition":"Baseline"}. This record does not assert independent reproduction.

### GPT-4

VGR: 0. Result ID: 6a2a76b8fbc373592c27af8b.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f38290966ade919c1a74.md>)

Recorded evaluation configuration: {"n (Sample Size)":"300","Prompt Condition":"Baseline"}. This record does not assert independent reproduction.

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