# Constitutional AI Alignment on HH-RLHF (test)

Source: [Constitutional AI Alignment on HH-RLHF (test)](<https://www.sota2.com/research/sota/constitutional-ai-alignment-on-hh-rlhf-test>)

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

Data updated: 2026-02-26T05:48:36.888Z

Evaluation of model performance regarding Constitutional AI alignment principles using the test split of the HH-RLHF dataset, measured by Likert scores and principle violation 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 | Constitutional AI Alignment |
| Dataset (including any stated split) | HH-RLHF (test) |
| Metric | Likert Score Ranking |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 6 of 6 |
| 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

- [Likert Score Ranking](<https://www.sota2.com/research/sota/constitutional-ai-alignment-on-hh-rlhf-test.md?metric=Likert%20Score%20Ranking>): Higher is better; unit: SCALAR.

- [Principle Violation Rate](<https://www.sota2.com/research/sota/constitutional-ai-alignment-on-hh-rlhf-test.md?metric=Principle%20Violation%20Rate>): Lower is better; unit: PERCENT.

- [Delta](<https://www.sota2.com/research/sota/constitutional-ai-alignment-on-hh-rlhf-test.md?metric=Delta>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Likert Score Ranking | Result date | Report paper |
| --- | --- | --- | --- |
| Reflect | 4.596 | Not recorded | 6978471ca2b28e761d87c9fd |
| Reflect | 4.406 | Not recorded | 6978471ca2b28e761d87c9fd |
| CCBase | 4.224 | Not recorded | 6978471ca2b28e761d87c9fd |
| Reflect | 4.179 | Not recorded | 6978471ca2b28e761d87c9fd |
| CCBase | 3.961 | Not recorded | 6978471ca2b28e761d87c9fd |
| CCBase | 3.944 | Not recorded | 6978471ca2b28e761d87c9fd |

### Reflect

Likert Score Ranking: 4.596. Result ID: 69966b215698665ec4b1d60b.

[Reporting paper](<https://www.sota2.com/research/paper/6978471ca2b28e761d87c9fd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6978471ca2b28e761d87c9fd.md>)

Recorded evaluation configuration: {"Backbone Model":"GPT-4.1-mini"}. This record does not assert independent reproduction.

### Reflect

Likert Score Ranking: 4.406. Result ID: 69966b215698665ec4b1d60d.

[Reporting paper](<https://www.sota2.com/research/paper/6978471ca2b28e761d87c9fd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6978471ca2b28e761d87c9fd.md>)

Recorded evaluation configuration: {"Backbone Model":"Claude-3.5-Haiku"}. This record does not assert independent reproduction.

### CCBase

Likert Score Ranking: 4.224. Result ID: 69966b215698665ec4b1d60a.

[Reporting paper](<https://www.sota2.com/research/paper/6978471ca2b28e761d87c9fd.md>)

Recorded evaluation configuration: {"Backbone Model":"GPT-4.1-mini"}. This record does not assert independent reproduction.

### Reflect

Likert Score Ranking: 4.179. Result ID: 69966b215698665ec4b1d60f.

[Reporting paper](<https://www.sota2.com/research/paper/6978471ca2b28e761d87c9fd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6978471ca2b28e761d87c9fd.md>)

Recorded evaluation configuration: {"Backbone Model":"Mistral-7B"}. This record does not assert independent reproduction.

### CCBase

Likert Score Ranking: 3.961. Result ID: 69966b215698665ec4b1d60e.

[Reporting paper](<https://www.sota2.com/research/paper/6978471ca2b28e761d87c9fd.md>)

Recorded evaluation configuration: {"Backbone Model":"Mistral-7B"}. This record does not assert independent reproduction.

### CCBase

Likert Score Ranking: 3.944. Result ID: 69966b215698665ec4b1d60c.

[Reporting paper](<https://www.sota2.com/research/paper/6978471ca2b28e761d87c9fd.md>)

Recorded evaluation configuration: {"Backbone Model":"Claude-3.5-Haiku"}. This record does not assert independent reproduction.

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