# LLM agent alignment evaluation on 1000 prompts (test)

Source: [LLM agent alignment evaluation on 1000 prompts (test)](<https://www.sota2.com/research/sota/llm-agent-alignment-evaluation-on-1000-prompts-test>)

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

Data updated: 2026-04-21T10:35:43.441Z

Evaluation of LLM agent alignment capabilities using the '1000 prompts (test)' dataset, measured by usefulness and neutrality scores.

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 | LLM agent alignment evaluation |
| Dataset (including any stated split) | 1000 prompts (test) |
| Metric | Usefulness Score |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 2 of 2 |
| 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

- [Usefulness Score](<https://www.sota2.com/research/sota/llm-agent-alignment-evaluation-on-1000-prompts-test.md?metric=Usefulness%20Score>): Higher is better; unit: SCALAR.

- [Neutrality Score](<https://www.sota2.com/research/sota/llm-agent-alignment-evaluation-on-1000-prompts-test.md?metric=Neutrality%20Score>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Usefulness Score | Result date | Report paper |
| --- | --- | --- | --- |
| LLM Default | 1 | Not recorded | 69e704a67d0172140ec7ee12 |
| LLM DReST | 1 | Not recorded | 69e704a67d0172140ec7ee12 |

### LLM Default

Usefulness Score: 1. Result ID: 69e751c247f34ce924ce4612.

[Reporting paper](<https://www.sota2.com/research/paper/69e704a67d0172140ec7ee12.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa725a1e209131cea55df7.md>)

Recorded evaluation configuration: {"Training duration (meta-episodes)":"1200","Number of random seeds":"5"}. This record does not assert independent reproduction.

### LLM DReST

Usefulness Score: 1. Result ID: 69e751c247f34ce924ce4613.

[Reporting paper](<https://www.sota2.com/research/paper/69e704a67d0172140ec7ee12.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69e704a67d0172140ec7ee12.md>)

Recorded evaluation configuration: {"Training duration (meta-episodes)":"1200","Number of random seeds":"5"}. This record does not assert independent reproduction.

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