# Multilingual text query safety on RTP-LX

Source: [Multilingual text query safety on RTP-LX](<https://www.sota2.com/research/sota/multilingual-text-query-safety-on-rtp-lx>)

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

Data updated: 2026-06-23T11:52:00.575Z

Evaluation of model performance on multilingual text query safety using classification metrics (Accuracy, Precision, Recall, F1) on the RTP-LX dataset.

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 | Multilingual text query safety |
| Dataset (including any stated split) | RTP-LX |
| Metric | Accuracy |
| Unit | PERCENT |
| 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

- [Accuracy](<https://www.sota2.com/research/sota/multilingual-text-query-safety-on-rtp-lx.md?metric=Accuracy>): Higher is better; unit: PERCENT.

- [Precision](<https://www.sota2.com/research/sota/multilingual-text-query-safety-on-rtp-lx.md?metric=Precision>): Higher is better; unit: PERCENT.

- [Recall](<https://www.sota2.com/research/sota/multilingual-text-query-safety-on-rtp-lx.md?metric=Recall>): Higher is better; unit: PERCENT.

- [F1 Score](<https://www.sota2.com/research/sota/multilingual-text-query-safety-on-rtp-lx.md?metric=F1%20Score>): Higher is better; unit: PERCENT.

- [Average F1 Score](<https://www.sota2.com/research/sota/multilingual-text-query-safety-on-rtp-lx.md?metric=Average%20F1%20Score>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| SingGuard-8B | 84.65 | Not recorded | 6a3a13aa206542251d7c7796 |
| SingGuard-4B | 81.7 | Not recorded | 6a3a13aa206542251d7c7796 |
| Qwen3Guard-8B-strict | 80.24 | Not recorded | 6a3a13aa206542251d7c7796 |
| SingGuard-2B | 80.1 | Not recorded | 6a3a13aa206542251d7c7796 |
| GraniteGuardian | 79.95 | Not recorded | 6a3a13aa206542251d7c7796 |
| YuFeng-XGuard-Reason-8B | 79.15 | Not recorded | 6a3a13aa206542251d7c7796 |
| Gemini3-Pro | 71.18 | Not recorded | 6a3a13aa206542251d7c7796 |
| GuardReasoner-VL-7B | 66.08 | Not recorded | 6a3a13aa206542251d7c7796 |
| GPT-5.1 | 57.41 | Not recorded | 6a3a13aa206542251d7c7796 |
| WildGuard | 52.85 | Not recorded | 6a3a13aa206542251d7c7796 |
| Qwen3Guard-8B-loose | 50.65 | Not recorded | 6a3a13aa206542251d7c7796 |
| Llama Guard 3 | 45.71 | Not recorded | 6a3a13aa206542251d7c7796 |
| ShieldGemma-9B | 44.54 | Not recorded | 6a3a13aa206542251d7c7796 |
| Qwen3-VL-235B | 35.7 | Not recorded | 6a3a13aa206542251d7c7796 |

### SingGuard-8B

Accuracy: 84.65. Result ID: 6a3a7347fbc373592c38383a.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13aa206542251d7c7796.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a3a13aa206542251d7c7796.md>)

[Implementation](<https://github.com/inclusionAI/Sing-Guard>)

Recorded evaluation configuration: {"Source":"Open-Source Models","Model Scale":"8B"}. This record does not assert independent reproduction.

### SingGuard-4B

Accuracy: 81.7. Result ID: 6a3a7347fbc373592c383839.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13aa206542251d7c7796.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a3a13aa206542251d7c7796.md>)

[Implementation](<https://github.com/inclusionAI/Sing-Guard>)

Recorded evaluation configuration: {"Source":"Open-Source Models","Model Scale":"4B"}. This record does not assert independent reproduction.

### Qwen3Guard-8B-strict

Accuracy: 80.24. Result ID: 6a3a7347fbc373592c383831.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13aa206542251d7c7796.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c351e209131cea8d09c.md>)

Recorded evaluation configuration: {"Source":"Open-Source Models","Model Scale":"8B","Safety Policy":"strict"}. This record does not assert independent reproduction.

### SingGuard-2B

Accuracy: 80.1. Result ID: 6a3a7347fbc373592c383838.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13aa206542251d7c7796.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a3a13aa206542251d7c7796.md>)

[Implementation](<https://github.com/inclusionAI/Sing-Guard>)

Recorded evaluation configuration: {"Source":"Open-Source Models","Model Scale":"2B"}. This record does not assert independent reproduction.

### GraniteGuardian

Accuracy: 79.95. Result ID: 6a3a7347fbc373592c383836.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13aa206542251d7c7796.md>)

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

Recorded evaluation configuration: {"Source":"Open-Source Models"}. This record does not assert independent reproduction.

### YuFeng-XGuard-Reason-8B

Accuracy: 79.15. Result ID: 6a3a7347fbc373592c383832.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13aa206542251d7c7796.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69730068b2be0ed85a78c8a4.md>)

Recorded evaluation configuration: {"Source":"Open-Source Models","Model Scale":"8B"}. This record does not assert independent reproduction.

### Gemini3-Pro

Accuracy: 71.18. Result ID: 6a3a7347fbc373592c38382d.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13aa206542251d7c7796.md>)

Recorded evaluation configuration: {"Source":"Closed-Source Models"}. This record does not assert independent reproduction.

### GuardReasoner-VL-7B

Accuracy: 66.08. Result ID: 6a3a7347fbc373592c383833.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13aa206542251d7c7796.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa789e1e209131cea7963f.md>)

Recorded evaluation configuration: {"Source":"Open-Source Models","Model Scale":"7B"}. This record does not assert independent reproduction.

### GPT-5.1

Accuracy: 57.41. Result ID: 6a3a7347fbc373592c38382e.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13aa206542251d7c7796.md>)

Recorded evaluation configuration: {"Source":"Closed-Source Models"}. This record does not assert independent reproduction.

### WildGuard

Accuracy: 52.85. Result ID: 6a3a7347fbc373592c383835.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13aa206542251d7c7796.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76291e209131cea6b9a6.md>)

Recorded evaluation configuration: {"Source":"Open-Source Models"}. This record does not assert independent reproduction.

### Qwen3Guard-8B-loose

Accuracy: 50.65. Result ID: 6a3a7347fbc373592c383830.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13aa206542251d7c7796.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c351e209131cea8d09c.md>)

Recorded evaluation configuration: {"Source":"Open-Source Models","Model Scale":"8B","Safety Policy":"loose"}. This record does not assert independent reproduction.

### Llama Guard 3

Accuracy: 45.71. Result ID: 6a3a7347fbc373592c383834.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13aa206542251d7c7796.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa717b1e209131cea5186d.md>)

Recorded evaluation configuration: {"Source":"Open-Source Models"}. This record does not assert independent reproduction.

### ShieldGemma-9B

Accuracy: 44.54. Result ID: 6a3a7347fbc373592c383837.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13aa206542251d7c7796.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74661e209131cea617ba.md>)

Recorded evaluation configuration: {"Source":"Open-Source Models","Model Scale":"9B"}. This record does not assert independent reproduction.

### Qwen3-VL-235B

Accuracy: 35.7. Result ID: 6a3a7347fbc373592c38382f.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13aa206542251d7c7796.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6927db54d6b0dd3342930f66.md>)

Recorded evaluation configuration: {"Source":"Open-Source Models","Model Scale":"235B"}. This record does not assert independent reproduction.

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