# Safety Classification on XSTestResponse

Source: [Safety Classification on XSTestResponse](<https://www.sota2.com/research/sota/safety-classification-on-xstestresponse>)

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

Data updated: 2026-02-26T05:43:33.079Z

Evaluation of model performance on the Safety Classification task using the XSTestResponse dataset, measured by F1 score.

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 | Safety Classification |
| Dataset (including any stated split) | XSTestResponse |
| Metric | F1 Score |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 14 of 14 |
| 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

- [F1 Score](<https://www.sota2.com/research/sota/safety-classification-on-xstestresponse.md?metric=F1%20Score>): Higher is better; unit: SCALAR.

## Recorded results

| Method | F1 Score | Result date | Report paper |
| --- | --- | --- | --- |
| AprielGuard | 0.96 | Not recorded | 694b737a53fdd5c3617399ba |
| AprielGuard | 0.95 | Not recorded | 694b737a53fdd5c3617399ba |
| Qwen3Guard | 0.94 | Not recorded | 694b737a53fdd5c3617399ba |
| Qwen3Guard | 0.92 | Not recorded | 694b737a53fdd5c3617399ba |
| Llama Guard 2 | 0.91 | Not recorded | 694b737a53fdd5c3617399ba |
| Llama Guard 3 | 0.9 | Not recorded | 694b737a53fdd5c3617399ba |
| Llama Guard 4 | 0.9 | Not recorded | 694b737a53fdd5c3617399ba |
| IBM Granite Guardian 3.3 | 0.9 | Not recorded | 694b737a53fdd5c3617399ba |
| IBM Granite Guardian 3.2 | 0.89 | Not recorded | 694b737a53fdd5c3617399ba |
| gpt-oss-safeguard | 0.89 | Not recorded | 694b737a53fdd5c3617399ba |
| IBM Granite Guardian 3.1 | 0.88 | Not recorded | 694b737a53fdd5c3617399ba |
| IBM Granite Guardian 3.3 | 0.86 | Not recorded | 694b737a53fdd5c3617399ba |
| IBM Granite Guardian 3.2 | 0.85 | Not recorded | 694b737a53fdd5c3617399ba |
| ShieldGemma | 0.83 | Not recorded | 694b737a53fdd5c3617399ba |

### AprielGuard

F1 Score: 0.96. Result ID: 6996b3765698665ec4fdd5cd.

[Reporting paper](<https://www.sota2.com/research/paper/694b737a53fdd5c3617399ba.md>)

[Method origin paper](<https://www.sota2.com/research/paper/694b737a53fdd5c3617399ba.md>)

Recorded evaluation configuration: {"Size":"8B","Reasoning":true}. This record does not assert independent reproduction.

### AprielGuard

F1 Score: 0.95. Result ID: 6996b3765698665ec4fdd5ca.

[Reporting paper](<https://www.sota2.com/research/paper/694b737a53fdd5c3617399ba.md>)

[Method origin paper](<https://www.sota2.com/research/paper/694b737a53fdd5c3617399ba.md>)

Recorded evaluation configuration: {"Size":"8B","Reasoning":false}. This record does not assert independent reproduction.

### Qwen3Guard

F1 Score: 0.94. Result ID: 6996b3765698665ec4fdd5c8.

[Reporting paper](<https://www.sota2.com/research/paper/694b737a53fdd5c3617399ba.md>)

Recorded evaluation configuration: {"Size":"8B","Mode":"loose","Reasoning":false}. This record does not assert independent reproduction.

### Qwen3Guard

F1 Score: 0.92. Result ID: 6996b3765698665ec4fdd5c9.

[Reporting paper](<https://www.sota2.com/research/paper/694b737a53fdd5c3617399ba.md>)

Recorded evaluation configuration: {"Size":"8B","Mode":"strict","Reasoning":false}. This record does not assert independent reproduction.

### Llama Guard 2

F1 Score: 0.91. Result ID: 6996b3765698665ec4fdd5c1.

[Reporting paper](<https://www.sota2.com/research/paper/694b737a53fdd5c3617399ba.md>)

Recorded evaluation configuration: {"Reasoning":false}. This record does not assert independent reproduction.

### Llama Guard 3

F1 Score: 0.9. Result ID: 6996b3765698665ec4fdd5c2.

[Reporting paper](<https://www.sota2.com/research/paper/694b737a53fdd5c3617399ba.md>)

Recorded evaluation configuration: {"Reasoning":false}. This record does not assert independent reproduction.

### Llama Guard 4

F1 Score: 0.9. Result ID: 6996b3765698665ec4fdd5c3.

[Reporting paper](<https://www.sota2.com/research/paper/694b737a53fdd5c3617399ba.md>)

Recorded evaluation configuration: {"Reasoning":false}. This record does not assert independent reproduction.

### IBM Granite Guardian 3.3

F1 Score: 0.9. Result ID: 6996b3765698665ec4fdd5c7.

[Reporting paper](<https://www.sota2.com/research/paper/694b737a53fdd5c3617399ba.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76331e209131cea6bcd3.md>)

Recorded evaluation configuration: {"Size":"8B","Reasoning":false}. This record does not assert independent reproduction.

### IBM Granite Guardian 3.2

F1 Score: 0.89. Result ID: 6996b3765698665ec4fdd5c6.

[Reporting paper](<https://www.sota2.com/research/paper/694b737a53fdd5c3617399ba.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76331e209131cea6bcd3.md>)

Recorded evaluation configuration: {"Size":"5B","Reasoning":false}. This record does not assert independent reproduction.

### gpt-oss-safeguard

F1 Score: 0.89. Result ID: 6996b3765698665ec4fdd5cc.

[Reporting paper](<https://www.sota2.com/research/paper/694b737a53fdd5c3617399ba.md>)

Recorded evaluation configuration: {"Size":"20B","Reasoning":true}. This record does not assert independent reproduction.

### IBM Granite Guardian 3.1

F1 Score: 0.88. Result ID: 6996b3765698665ec4fdd5c4.

[Reporting paper](<https://www.sota2.com/research/paper/694b737a53fdd5c3617399ba.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76331e209131cea6bcd3.md>)

Recorded evaluation configuration: {"Size":"2B","Reasoning":false}. This record does not assert independent reproduction.

### IBM Granite Guardian 3.3

F1 Score: 0.86. Result ID: 6996b3765698665ec4fdd5cb.

[Reporting paper](<https://www.sota2.com/research/paper/694b737a53fdd5c3617399ba.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76331e209131cea6bcd3.md>)

Recorded evaluation configuration: {"Size":"8B","Reasoning":true}. This record does not assert independent reproduction.

### IBM Granite Guardian 3.2

F1 Score: 0.85. Result ID: 6996b3765698665ec4fdd5c5.

[Reporting paper](<https://www.sota2.com/research/paper/694b737a53fdd5c3617399ba.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76331e209131cea6bcd3.md>)

Recorded evaluation configuration: {"Size":"3B","Reasoning":false}. This record does not assert independent reproduction.

### ShieldGemma

F1 Score: 0.83. Result ID: 6996b3765698665ec4fdd5c0.

[Reporting paper](<https://www.sota2.com/research/paper/694b737a53fdd5c3617399ba.md>)

Recorded evaluation configuration: {"Size":"9B","Reasoning":false}. This record does not assert independent reproduction.

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