# Prompt Injection Detection on AgentDojo (Precision/Recall Metrics) Source: [Prompt Injection Detection on AgentDojo (Precision/Recall Metrics)]() Collection index: [llms.txt]() Data updated: 2026-06-12T06:58:22.135Z Evaluation of model performance on Prompt Injection Detection using the AgentDojo dataset, specifically measuring Source (Src) and Instruction (Ins) Precision and Recall. 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 | Prompt Injection Detection | | Dataset (including any stated split) | AgentDojo | | Metric | Source Precision | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 30 of 30 | | 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 - [Source Precision](): Higher is better; unit: PERCENT. - [Source Recall](): Higher is better; unit: PERCENT. - [Instruction Precision](): Higher is better; unit: PERCENT. - [Instruction Recall](): Higher is better; unit: PERCENT. ## Recorded results | Method | Source Precision | Result date | Report paper | | --- | --- | --- | --- | | PI-Hunter | 92.1 | Not recorded | 6a2b9278d33dfb23e783af62 | | PI-Hunter | 92 | Not recorded | 6a2b9278d33dfb23e783af62 | | PI-Hunter | 90.7 | Not recorded | 6a2b9278d33dfb23e783af62 | | PI-Hunter | 89.1 | Not recorded | 6a2b9278d33dfb23e783af62 | | PI-Hunter | 87.7 | Not recorded | 6a2b9278d33dfb23e783af62 | | PI-Hunter | 87.7 | Not recorded | 6a2b9278d33dfb23e783af62 | | PI-Hunter | 87.3 | Not recorded | 6a2b9278d33dfb23e783af62 | | PI-Hunter | 87 | Not recorded | 6a2b9278d33dfb23e783af62 | | Baseline | 85.4 | Not recorded | 6a2b9278d33dfb23e783af62 | | Baseline | 83.9 | Not recorded | 6a2b9278d33dfb23e783af62 | | PI-Hunter | 83.4 | Not recorded | 6a2b9278d33dfb23e783af62 | | Baseline | 82.9 | Not recorded | 6a2b9278d33dfb23e783af62 | | Baseline | 82.6 | Not recorded | 6a2b9278d33dfb23e783af62 | | PI-Hunter | 81.2 | Not recorded | 6a2b9278d33dfb23e783af62 | | PI-Hunter | 80.3 | Not recorded | 6a2b9278d33dfb23e783af62 | | PI-Hunter | 79.6 | Not recorded | 6a2b9278d33dfb23e783af62 | | Baseline | 77.3 | Not recorded | 6a2b9278d33dfb23e783af62 | | PI-Hunter | 77.3 | Not recorded | 6a2b9278d33dfb23e783af62 | | PI-Hunter | 75.4 | Not recorded | 6a2b9278d33dfb23e783af62 | | PI-Hunter | 73.4 | Not recorded | 6a2b9278d33dfb23e783af62 | | Baseline | 73.4 | Not recorded | 6a2b9278d33dfb23e783af62 | | Baseline | 71.9 | Not recorded | 6a2b9278d33dfb23e783af62 | | Baseline | 70.3 | Not recorded | 6a2b9278d33dfb23e783af62 | | Baseline | 59.4 | Not recorded | 6a2b9278d33dfb23e783af62 | | Baseline | 59 | Not recorded | 6a2b9278d33dfb23e783af62 | | Baseline | 57.6 | Not recorded | 6a2b9278d33dfb23e783af62 | | Baseline | 55.8 | Not recorded | 6a2b9278d33dfb23e783af62 | | Baseline | 50.5 | Not recorded | 6a2b9278d33dfb23e783af62 | | Baseline | 49 | Not recorded | 6a2b9278d33dfb23e783af62 | | Baseline | 36 | Not recorded | 6a2b9278d33dfb23e783af62 | ### PI-Hunter Source Precision: 92.1. Result ID: 6a2bae07fbc373592c297e88. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Gemini-3.1-pro","Attack":"important\_inst"}. This record does not assert independent reproduction. ### PI-Hunter Source Precision: 92. Result ID: 6a2bae07fbc373592c297e84. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Gemini-3.1-pro","Attack":"ignore\_previous"}. This record does not assert independent reproduction. ### PI-Hunter Source Precision: 90.7. Result ID: 6a2bae07fbc373592c297e86. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Gemini-3.1-pro","Attack":"system\_msg"}. This record does not assert independent reproduction. ### PI-Hunter Source Precision: 89.1. Result ID: 6a2bae07fbc373592c297e98. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Claude-4.6-sonnet","Attack":"ignore\_previous"}. This record does not assert independent reproduction. ### PI-Hunter Source Precision: 87.7. Result ID: 6a2bae07fbc373592c297e96. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Claude-4.6-sonnet","Attack":"direct"}. This record does not assert independent reproduction. ### PI-Hunter Source Precision: 87.7. Result ID: 6a2bae07fbc373592c297e9a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Claude-4.6-sonnet","Attack":"system\_msg"}. This record does not assert independent reproduction. ### PI-Hunter Source Precision: 87.3. Result ID: 6a2bae07fbc373592c297e82. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Gemini-3.1-pro","Attack":"direct"}. This record does not assert independent reproduction. ### PI-Hunter Source Precision: 87. Result ID: 6a2bae07fbc373592c297e9c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Claude-4.6-sonnet","Attack":"important\_inst"}. This record does not assert independent reproduction. ### Baseline Source Precision: 85.4. Result ID: 6a2bae07fbc373592c297e95. [Reporting paper]() Recorded evaluation configuration: {"Model":"Claude-4.6-sonnet","Attack":"direct"}. This record does not assert independent reproduction. ### Baseline Source Precision: 83.9. Result ID: 6a2bae07fbc373592c297e9b. [Reporting paper]() Recorded evaluation configuration: {"Model":"Claude-4.6-sonnet","Attack":"important\_inst"}. This record does not assert independent reproduction. ### PI-Hunter Source Precision: 83.4. Result ID: 6a2bae07fbc373592c297e8c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"GPT-5.4-mini","Attack":"direct"}. This record does not assert independent reproduction. ### Baseline Source Precision: 82.9. Result ID: 6a2bae07fbc373592c297e97. [Reporting paper]() Recorded evaluation configuration: {"Model":"Claude-4.6-sonnet","Attack":"ignore\_previous"}. This record does not assert independent reproduction. ### Baseline Source Precision: 82.6. Result ID: 6a2bae07fbc373592c297e99. [Reporting paper]() Recorded evaluation configuration: {"Model":"Claude-4.6-sonnet","Attack":"system\_msg"}. This record does not assert independent reproduction. ### PI-Hunter Source Precision: 81.2. Result ID: 6a2bae07fbc373592c297e8e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"GPT-5.4-mini","Attack":"ignore\_previous"}. This record does not assert independent reproduction. ### PI-Hunter Source Precision: 80.3. Result ID: 6a2bae07fbc373592c297e94. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"GPT-5.4-mini","Attack":"agentvigil"}. This record does not assert independent reproduction. ### PI-Hunter Source Precision: 79.6. Result ID: 6a2bae07fbc373592c297e8a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Gemini-3.1-pro","Attack":"agentvigil"}. This record does not assert independent reproduction. ### Baseline Source Precision: 77.3. Result ID: 6a2bae07fbc373592c297e8b. [Reporting paper]() Recorded evaluation configuration: {"Model":"GPT-5.4-mini","Attack":"direct"}. This record does not assert independent reproduction. ### PI-Hunter Source Precision: 77.3. Result ID: 6a2bae07fbc373592c297e9e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Claude-4.6-sonnet","Attack":"agentvigil"}. This record does not assert independent reproduction. ### PI-Hunter Source Precision: 75.4. Result ID: 6a2bae07fbc373592c297e90. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"GPT-5.4-mini","Attack":"system\_msg"}. This record does not assert independent reproduction. ### PI-Hunter Source Precision: 73.4. Result ID: 6a2bae07fbc373592c297e92. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"GPT-5.4-mini","Attack":"important\_inst"}. This record does not assert independent reproduction. ### Baseline Source Precision: 73.4. Result ID: 6a2bae07fbc373592c297e9d. [Reporting paper]() Recorded evaluation configuration: {"Model":"Claude-4.6-sonnet","Attack":"agentvigil"}. This record does not assert independent reproduction. ### Baseline Source Precision: 71.9. Result ID: 6a2bae07fbc373592c297e8d. [Reporting paper]() Recorded evaluation configuration: {"Model":"GPT-5.4-mini","Attack":"ignore\_previous"}. This record does not assert independent reproduction. ### Baseline Source Precision: 70.3. Result ID: 6a2bae07fbc373592c297e93. [Reporting paper]() Recorded evaluation configuration: {"Model":"GPT-5.4-mini","Attack":"agentvigil"}. This record does not assert independent reproduction. ### Baseline Source Precision: 59.4. Result ID: 6a2bae07fbc373592c297e8f. [Reporting paper]() Recorded evaluation configuration: {"Model":"GPT-5.4-mini","Attack":"system\_msg"}. This record does not assert independent reproduction. ### Baseline Source Precision: 59. Result ID: 6a2bae07fbc373592c297e91. [Reporting paper]() Recorded evaluation configuration: {"Model":"GPT-5.4-mini","Attack":"important\_inst"}. This record does not assert independent reproduction. ### Baseline Source Precision: 57.6. Result ID: 6a2bae07fbc373592c297e85. [Reporting paper]() Recorded evaluation configuration: {"Model":"Gemini-3.1-pro","Attack":"system\_msg"}. This record does not assert independent reproduction. ### Baseline Source Precision: 55.8. Result ID: 6a2bae07fbc373592c297e89. [Reporting paper]() Recorded evaluation configuration: {"Model":"Gemini-3.1-pro","Attack":"agentvigil"}. This record does not assert independent reproduction. ### Baseline Source Precision: 50.5. Result ID: 6a2bae07fbc373592c297e87. [Reporting paper]() Recorded evaluation configuration: {"Model":"Gemini-3.1-pro","Attack":"important\_inst"}. This record does not assert independent reproduction. ### Baseline Source Precision: 49. Result ID: 6a2bae07fbc373592c297e81. [Reporting paper]() Recorded evaluation configuration: {"Model":"Gemini-3.1-pro","Attack":"direct"}. This record does not assert independent reproduction. ### Baseline Source Precision: 36. Result ID: 6a2bae07fbc373592c297e83. [Reporting paper]() Recorded evaluation configuration: {"Model":"Gemini-3.1-pro","Attack":"ignore\_previous"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.