# Safe and Helpful Response Generation on SafeRLHF-30K (test)

Source: [Safe and Helpful Response Generation on SafeRLHF-30K (test)](<https://www.sota2.com/research/sota/safe-and-helpful-response-generation-on-saferlhf-30k-test>)

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

Data updated: 2026-03-05T06:47:15.737Z

Evaluation of model performance on generating responses that are both safe and helpful, using the test split of the SafeRLHF-30K dataset, measured by Safe Response Rate and Helpful Response Rate.

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 | Safe and Helpful Response Generation |
| Dataset (including any stated split) | SafeRLHF-30K (test) |
| Metric | Safe Response Rate |
| 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

- [Safe Response Rate](<https://www.sota2.com/research/sota/safe-and-helpful-response-generation-on-saferlhf-30k-test.md?metric=Safe%20Response%20Rate>): Higher is better; unit: PERCENT.

- [Helpful Response Rate](<https://www.sota2.com/research/sota/safe-and-helpful-response-generation-on-saferlhf-30k-test.md?metric=Helpful%20Response%20Rate>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Safe Response Rate | Result date | Report paper |
| --- | --- | --- | --- |
| SafeDPO | 94.3 | Not recorded | 68fa78d31e209131cea7a9d6 |
| SafeRLHF | 93.2 | Not recorded | 68fa78d31e209131cea7a9d6 |
| SFT | 43 | Not recorded | 68fa78d31e209131cea7a9d6 |

### SafeDPO

Safe Response Rate: 94.3. Result ID: 69a926926dbb7075b951e5d3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78d31e209131cea7a9d6.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78d31e209131cea7a9d6.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### SafeRLHF

Safe Response Rate: 93.2. Result ID: 69a926926dbb7075b951e5d2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78d31e209131cea7a9d6.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70e81e209131cea4ddd5.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### SFT

Safe Response Rate: 43. Result ID: 69a926926dbb7075b951e5d1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78d31e209131cea7a9d6.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

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