# Skill Invocation Safety Auditing on SIA-Bench Held-out indirect prompt injection Source: [Skill Invocation Safety Auditing on SIA-Bench Held-out indirect prompt injection]() Collection index: [llms.txt]() Data updated: 2026-04-14T08:52:43.502Z Evaluation of model safety and auditing performance concerning skill invocation when subjected to indirect prompt injection attacks, utilizing the SIA-Bench Held-out dataset. 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 | Skill Invocation Safety Auditing | | Dataset (including any stated split) | SIA-Bench Held-out indirect prompt injection | | Metric | HR-AUPRC | | 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 - [HR-AUPRC](): Higher is better; unit: PERCENT. - [Recall@10%](): Higher is better; unit: PERCENT. - [Precision@10%](): Higher is better; unit: PERCENT. - [Spearman Correlation](): Higher is better; unit: SCALAR. - [ECE](): Lower is better; unit: PERCENT. - [W-MAE](): Lower is better; unit: SCALAR. ## Recorded results | Method | HR-AUPRC | Result date | Report paper | | --- | --- | --- | --- | | Fusion | 43.9 | Not recorded | 69ddca377d0172140ec79b88 | | Contextual | 40.5 | Not recorded | 69ddca377d0172140ec79b88 | | Text-Only | 39.9 | Not recorded | 69ddca377d0172140ec79b88 | | Agent-Audit | 38 | Not recorded | 69ddca377d0172140ec79b88 | | No Audit | 36.4 | Not recorded | 69ddca377d0172140ec79b88 | | Static Prior | 33.9 | Not recorded | 69ddca377d0172140ec79b88 | | Denylist | 33.9 | Not recorded | 69ddca377d0172140ec79b88 | ### Fusion HR-AUPRC: 43.9. Result ID: 69de005647f34ce924c58cb0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Contextual HR-AUPRC: 40.5. Result ID: 69de005647f34ce924c58caf. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Text-Only HR-AUPRC: 39.9. Result ID: 69de005647f34ce924c58cae. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Agent-Audit HR-AUPRC: 38. Result ID: 69de005647f34ce924c58cad. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### No Audit HR-AUPRC: 36.4. Result ID: 69de005647f34ce924c58caa. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Static Prior HR-AUPRC: 33.9. Result ID: 69de005647f34ce924c58cab. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Denylist HR-AUPRC: 33.9. Result ID: 69de005647f34ce924c58cac. [Reporting paper]() 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.