# Safety Evaluation and Alignment on ATBench Family

Source: [Safety Evaluation and Alignment on ATBench Family](<https://www.sota2.com/research/sota/safety-evaluation-and-alignment-on-atbench-family>)

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

Data updated: 2026-05-29T10:13:36.085Z

Evaluation of model safety and alignment performance using the ATBench Family dataset, focusing on judgment expertise, rationale quality, and agentic safety monitoring capabilities.

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 Evaluation and Alignment |
| Dataset (including any stated split) | ATBench Family |
| Metric | Judgment Expertise |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 0 of 0 |
| 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

- [Judgment Expertise](<https://www.sota2.com/research/sota/safety-evaluation-and-alignment-on-atbench-family.md?metric=Judgment%20Expertise>): Higher is better; unit: PERCENT.

- [Coarse Judgment](<https://www.sota2.com/research/sota/safety-evaluation-and-alignment-on-atbench-family.md?metric=Coarse%20Judgment>): Higher is better; unit: PERCENT.

- [Fine-Grained Judgment](<https://www.sota2.com/research/sota/safety-evaluation-and-alignment-on-atbench-family.md?metric=Fine-Grained%20Judgment>): Higher is better; unit: PERCENT.

- [Judgment Rationale Quality](<https://www.sota2.com/research/sota/safety-evaluation-and-alignment-on-atbench-family.md?metric=Judgment%20Rationale%20Quality>): Higher is better; unit: PERCENT.

- [Safety-Aware Agentic Training Score](<https://www.sota2.com/research/sota/safety-evaluation-and-alignment-on-atbench-family.md?metric=Safety-Aware%20Agentic%20Training%20Score>): Higher is better; unit: PERCENT.

- [Online Agentic Safety Monitoring](<https://www.sota2.com/research/sota/safety-evaluation-and-alignment-on-atbench-family.md?metric=Online%20Agentic%20Safety%20Monitoring>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Judgment Expertise | Result date | Report paper |
| --- | --- | --- | --- |

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