# Behavioral Axis Detection on ACS In-distribution 10 axes (test) Source: [Behavioral Axis Detection on ACS In-distribution 10 axes (test)]() Collection index: [llms.txt]() Data updated: 2026-05-12T06:40:08.552Z Evaluation of model performance on detecting specific behavioral axes (e.g., refusal, math, bias, toxicity) using the test split of the ACS In-distribution 10 axes dataset, reporting MC Recall and BC AUROC for each axis. 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 | Behavioral Axis Detection | | Dataset (including any stated split) | ACS In-distribution 10 axes (test) | | Metric | Refusal MC Recall | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 7 of 7 | | 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 - [Refusal MC Recall](): Higher is better; unit: PERCENT. - [Refusal BC AUROC](): Higher is better; unit: PERCENT. - [Math MC Recall](): Higher is better; unit: PERCENT. - [Math BC AUROC](): Higher is better; unit: PERCENT. - [Science MC Recall](): Higher is better; unit: PERCENT. - [Science BC AUROC](): Higher is better; unit: PERCENT. - [Factual MC Recall](): Higher is better; unit: PERCENT. - [Factual BC AUROC](): Higher is better; unit: PERCENT. - [Sycophancy MC Recall](): Higher is better; unit: PERCENT. - [Sycophancy BC AUROC](): Higher is better; unit: PERCENT. - [Toxicity MC Recall](): Higher is better; unit: PERCENT. - [Toxicity BC AUROC](): Higher is better; unit: PERCENT. - [Sentiment MC Recall](): Higher is better; unit: PERCENT. - [Sentiment BC AUROC](): Higher is better; unit: PERCENT. - [Emotion MC Recall](): Higher is better; unit: PERCENT. - [Emotion BC AUROC](): Higher is better; unit: PERCENT. - [Bias (Gender) MC Recall](): Higher is better; unit: PERCENT. - [Bias (Gender) BC AUROC](): Higher is better; unit: PERCENT. - [Bias (Race) MC Recall](): Higher is better; unit: PERCENT. - [Bias (Race) BC AUROC](): Higher is better; unit: PERCENT. - [Mean MC Recall (10 Axes)](): Higher is better; unit: PERCENT. - [Mean BC AUROC (10 Axes)](): Higher is better; unit: PERCENT. ## Recorded results | Method | Refusal MC Recall | Result date | Report paper | | --- | --- | --- | --- | | ACS Behavioral Probe | 97 | Not recorded | 6a02b48edbc836b87d824cd0 | | ACS Behavioral Probe | 95 | Not recorded | 6a02b48edbc836b87d824cd0 | | ACS Behavioral Probe | 93 | Not recorded | 6a02b48edbc836b87d824cd0 | | ACS Behavioral Probe | 92 | Not recorded | 6a02b48edbc836b87d824cd0 | | ACS Behavioral Probe | 92 | Not recorded | 6a02b48edbc836b87d824cd0 | | ACS Behavioral Probe | 88 | Not recorded | 6a02b48edbc836b87d824cd0 | | ACS Behavioral Probe | 87 | Not recorded | 6a02b48edbc836b87d824cd0 | ### ACS Behavioral Probe Refusal MC Recall: 97. Result ID: 6a02ca7f47f34ce924e4f655. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target Model":"Mistral"}. This record does not assert independent reproduction. ### ACS Behavioral Probe Refusal MC Recall: 95. Result ID: 6a02ca7f47f34ce924e4f653. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target Model":"Llama"}. This record does not assert independent reproduction. ### ACS Behavioral Probe Refusal MC Recall: 93. Result ID: 6a02ca7f47f34ce924e4f658. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target Model":"LQMP (4-fam)"}. This record does not assert independent reproduction. ### ACS Behavioral Probe Refusal MC Recall: 92. Result ID: 6a02ca7f47f34ce924e4f656. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target Model":"Phi"}. This record does not assert independent reproduction. ### ACS Behavioral Probe Refusal MC Recall: 92. Result ID: 6a02ca7f47f34ce924e4f659. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target Model":"LQMPG (5-fam)"}. This record does not assert independent reproduction. ### ACS Behavioral Probe Refusal MC Recall: 88. Result ID: 6a02ca7f47f34ce924e4f654. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target Model":"Qwen"}. This record does not assert independent reproduction. ### ACS Behavioral Probe Refusal MC Recall: 87. Result ID: 6a02ca7f47f34ce924e4f657. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target Model":"Gemma"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.