# Harmful Content Detection on PHEME Known Attacks: DeepWordBug, TFAdjusted, TREPAT (test) Source: [Harmful Content Detection on PHEME Known Attacks: DeepWordBug, TFAdjusted, TREPAT (test)]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:32:22.880Z Evaluation of Harmful Content Detection performance on the PHEME dataset, specifically focusing on known adversarial attacks: DeepWordBug, TFAdjusted, and TREPAT, using the test split. 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 | Harmful Content Detection | | Dataset (including any stated split) | PHEME Known Attacks: DeepWordBug, TFAdjusted, TREPAT (test) | | Metric | Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 10 of 10 | | 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 - [Accuracy](): Higher is better; unit: PERCENT. - [Precision](): Higher is better; unit: PERCENT. - [Recall](): Higher is better; unit: PERCENT. - [F1-score](): Higher is better; unit: PERCENT. - [ASR](): Lower is better; unit: PERCENT. ## Recorded results | Method | Accuracy | Result date | Report paper | | --- | --- | --- | --- | | LLM-SGA/ARHOCD | 85.59 | Not recorded | 6948d05e423ea42265729da2 | | LLM-SGA/ARHOCD | 85.49 | Not recorded | 6948d05e423ea42265729da2 | | LLM-puri | 83.66 | Not recorded | 6948d05e423ea42265729da2 | | MRAT | 83.54 | Not recorded | 6948d05e423ea42265729da2 | | EnsSel | 81.81 | Not recorded | 6948d05e423ea42265729da2 | | AT | 81.59 | Not recorded | 6948d05e423ea42265729da2 | | Det&Res | 81.38 | Not recorded | 6948d05e423ea42265729da2 | | OutReg | 81.15 | Not recorded | 6948d05e423ea42265729da2 | | ARText | 80.95 | Not recorded | 6948d05e423ea42265729da2 | | ARDEL | 80.78 | Not recorded | 6948d05e423ea42265729da2 | ### LLM-SGA/ARHOCD Accuracy: 85.59. Result ID: 69994fa25698665ec4ba0f07. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"approximation":"Monte Carlo"}. This record does not assert independent reproduction. ### LLM-SGA/ARHOCD Accuracy: 85.49. Result ID: 69994fa25698665ec4ba0f08. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"approximation":"Lower-bound"}. This record does not assert independent reproduction. ### LLM-puri Accuracy: 83.66. Result ID: 69994fa25698665ec4ba0eff. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"mode":"Enhanced"}. This record does not assert independent reproduction. ### MRAT Accuracy: 83.54. Result ID: 69994fa25698665ec4ba0f03. [Reporting paper]() Recorded evaluation configuration: {"mode":"Enhanced"}. This record does not assert independent reproduction. ### EnsSel Accuracy: 81.81. Result ID: 69994fa25698665ec4ba0f06. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"mode":"Enhanced"}. This record does not assert independent reproduction. ### AT Accuracy: 81.59. Result ID: 69994fa25698665ec4ba0f02. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"mode":"Enhanced"}. This record does not assert independent reproduction. ### Det&Res Accuracy: 81.38. Result ID: 69994fa25698665ec4ba0f00. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"mode":"Enhanced"}. This record does not assert independent reproduction. ### OutReg Accuracy: 81.15. Result ID: 69994fa25698665ec4ba0f01. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"mode":"Enhanced"}. This record does not assert independent reproduction. ### ARText Accuracy: 80.95. Result ID: 69994fa25698665ec4ba0f04. [Reporting paper]() Recorded evaluation configuration: {"mode":"Enhanced"}. This record does not assert independent reproduction. ### ARDEL Accuracy: 80.78. Result ID: 69994fa25698665ec4ba0f05. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"mode":"Enhanced"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.