# Mirage Detection on VQA Mirage Detection Chest, Pathology, Natural, Document, Infographic (test) Source: [Mirage Detection on VQA Mirage Detection Chest, Pathology, Natural, Document, Infographic (test)]() Collection index: [llms.txt]() Data updated: 2026-06-02T08:11:01.807Z Evaluation of model performance for Mirage Detection on the test split of the VQA Mirage Detection dataset, reporting Accuracy, Mirage Rate (MR), and Macro F1 Score. 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 | Mirage Detection | | Dataset (including any stated split) | VQA Mirage Detection Chest, Pathology, Natural, Document, Infographic (test) | | Metric | Mirage Rate | | Unit | PERCENT | | Direction | Lower is better | | Results loaded | 36 of 36 | | 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 - [Mirage Rate](): Lower is better; unit: PERCENT. - [Accuracy](): Higher is better; unit: PERCENT. - [Macro F1 Score](): Higher is better; unit: PERCENT. ## Recorded results | Method | Mirage Rate | Result date | Report paper | | --- | --- | --- | --- | | TC-LIA + Ensemble | 2.7 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA + Ensemble | 2.8 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA + Ensemble | 2.8 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA + Ensemble | 3 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA + Ensemble | 3 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA + Ensemble | 3 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA + Ensemble | 3 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA + Ensemble | 3.1 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA + Ensemble | 3.1 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA + Ensemble | 3.2 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA + Ensemble | 3.2 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA + Ensemble | 3.3 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA | 3.4 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA | 3.4 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA | 3.4 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA | 3.4 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA | 3.4 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA | 3.4 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA | 3.4 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA | 3.4 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA | 3.4 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA | 3.4 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA | 3.4 | Not recorded | 6a1e63bbdbd7150506f852fd | | TC-LIA | 3.4 | Not recorded | 6a1e63bbdbd7150506f852fd | | Aya-Vision-32B | 21.7 | Not recorded | 6a1e63bbdbd7150506f852fd | | LLaVA-v1.6-34B | 24.9 | Not recorded | 6a1e63bbdbd7150506f852fd | | LLaMA-3.2-90B | 26 | Not recorded | 6a1e63bbdbd7150506f852fd | | InternVL3-38B | 27.1 | Not recorded | 6a1e63bbdbd7150506f852fd | | Qwen2.5-VL-32B | 57.4 | Not recorded | 6a1e63bbdbd7150506f852fd | | BLIP2-2.7B | 58 | Not recorded | 6a1e63bbdbd7150506f852fd | | Gemma3-4B | 58.8 | Not recorded | 6a1e63bbdbd7150506f852fd | | Phi-3.5-Vision | 61.9 | Not recorded | 6a1e63bbdbd7150506f852fd | | Qwen2.5-VL-72B | 63.6 | Not recorded | 6a1e63bbdbd7150506f852fd | | Qwen2.5-VL-7B | 63.6 | Not recorded | 6a1e63bbdbd7150506f852fd | | LLaVA-Next-110B | 63.6 | Not recorded | 6a1e63bbdbd7150506f852fd | | MiniCPM-V-2.6 | 66.6 | Not recorded | 6a1e63bbdbd7150506f852fd | ### TC-LIA + Ensemble Mirage Rate: 2.7. Result ID: 6a1e900afbc373592c1894b6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"LLaMA-3.2-90B","Detection Configuration":"TC-LIA + Ensemble"}. This record does not assert independent reproduction. ### TC-LIA + Ensemble Mirage Rate: 2.8. Result ID: 6a1e900afbc373592c1894b3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"Qwen2.5-VL-72B","Detection Configuration":"TC-LIA + Ensemble"}. This record does not assert independent reproduction. ### TC-LIA + Ensemble Mirage Rate: 2.8. Result ID: 6a1e900afbc373592c1894c8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"InternVL3-38B","Detection Configuration":"TC-LIA + Ensemble"}. This record does not assert independent reproduction. ### TC-LIA + Ensemble Mirage Rate: 3. Result ID: 6a1e900afbc373592c1894b0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"Qwen2.5-VL-32B","Detection Configuration":"TC-LIA + Ensemble"}. This record does not assert independent reproduction. ### TC-LIA + Ensemble Mirage Rate: 3. Result ID: 6a1e900afbc373592c1894c5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"LLaVA-Next-110B","Detection Configuration":"TC-LIA + Ensemble"}. This record does not assert independent reproduction. ### TC-LIA + Ensemble Mirage Rate: 3. Result ID: 6a1e900afbc373592c1894ce. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"Phi-3.5-Vision","Detection Configuration":"TC-LIA + Ensemble"}. This record does not assert independent reproduction. ### TC-LIA + Ensemble Mirage Rate: 3. Result ID: 6a1e900afbc373592c1894d1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"BLIP2-2.7B","Detection Configuration":"TC-LIA + Ensemble"}. This record does not assert independent reproduction. ### TC-LIA + Ensemble Mirage Rate: 3.1. Result ID: 6a1e900afbc373592c1894bf. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"Gemma3-4B","Detection Configuration":"TC-LIA + Ensemble"}. This record does not assert independent reproduction. ### TC-LIA + Ensemble Mirage Rate: 3.1. Result ID: 6a1e900afbc373592c1894cb. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"MiniCPM-V-2.6","Detection Configuration":"TC-LIA + Ensemble"}. This record does not assert independent reproduction. ### TC-LIA + Ensemble Mirage Rate: 3.2. Result ID: 6a1e900afbc373592c1894b9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"Aya-Vision-32B","Detection Configuration":"TC-LIA + Ensemble"}. This record does not assert independent reproduction. ### TC-LIA + Ensemble Mirage Rate: 3.2. Result ID: 6a1e900afbc373592c1894bc. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"Qwen2.5-VL-7B","Detection Configuration":"TC-LIA + Ensemble"}. This record does not assert independent reproduction. ### TC-LIA + Ensemble Mirage Rate: 3.3. Result ID: 6a1e900afbc373592c1894c2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"LLaVA-v1.6-34B","Detection Configuration":"TC-LIA + Ensemble"}. This record does not assert independent reproduction. ### TC-LIA Mirage Rate: 3.4. Result ID: 6a1e900afbc373592c1894af. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"Qwen2.5-VL-32B","Detection Configuration":"TC-LIA Only"}. This record does not assert independent reproduction. ### TC-LIA Mirage Rate: 3.4. Result ID: 6a1e900afbc373592c1894b2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"Qwen2.5-VL-72B","Detection Configuration":"TC-LIA Only"}. This record does not assert independent reproduction. ### TC-LIA Mirage Rate: 3.4. Result ID: 6a1e900afbc373592c1894b5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"LLaMA-3.2-90B","Detection Configuration":"TC-LIA Only"}. This record does not assert independent reproduction. ### TC-LIA Mirage Rate: 3.4. Result ID: 6a1e900afbc373592c1894b8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"Aya-Vision-32B","Detection Configuration":"TC-LIA Only"}. This record does not assert independent reproduction. ### TC-LIA Mirage Rate: 3.4. Result ID: 6a1e900afbc373592c1894bb. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"Qwen2.5-VL-7B","Detection Configuration":"TC-LIA Only"}. This record does not assert independent reproduction. ### TC-LIA Mirage Rate: 3.4. Result ID: 6a1e900afbc373592c1894be. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"Gemma3-4B","Detection Configuration":"TC-LIA Only"}. This record does not assert independent reproduction. ### TC-LIA Mirage Rate: 3.4. Result ID: 6a1e900afbc373592c1894c1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"LLaVA-v1.6-34B","Detection Configuration":"TC-LIA Only"}. This record does not assert independent reproduction. ### TC-LIA Mirage Rate: 3.4. Result ID: 6a1e900afbc373592c1894c4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"LLaVA-Next-110B","Detection Configuration":"TC-LIA Only"}. This record does not assert independent reproduction. ### TC-LIA Mirage Rate: 3.4. Result ID: 6a1e900afbc373592c1894c7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"InternVL3-38B","Detection Configuration":"TC-LIA Only"}. This record does not assert independent reproduction. ### TC-LIA Mirage Rate: 3.4. Result ID: 6a1e900afbc373592c1894ca. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"MiniCPM-V-2.6","Detection Configuration":"TC-LIA Only"}. This record does not assert independent reproduction. ### TC-LIA Mirage Rate: 3.4. Result ID: 6a1e900afbc373592c1894cd. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"Phi-3.5-Vision","Detection Configuration":"TC-LIA Only"}. This record does not assert independent reproduction. ### TC-LIA Mirage Rate: 3.4. Result ID: 6a1e900afbc373592c1894d0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLM Backbone":"BLIP2-2.7B","Detection Configuration":"TC-LIA Only"}. This record does not assert independent reproduction. ### Aya-Vision-32B Mirage Rate: 21.7. Result ID: 6a1e900afbc373592c1894b7. [Reporting paper]() Recorded evaluation configuration: {"Detection Configuration":"Base"}. This record does not assert independent reproduction. ### LLaVA-v1.6-34B Mirage Rate: 24.9. Result ID: 6a1e900afbc373592c1894c0. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Detection Configuration":"Base"}. This record does not assert independent reproduction. ### LLaMA-3.2-90B Mirage Rate: 26. Result ID: 6a1e900afbc373592c1894b4. [Reporting paper]() Recorded evaluation configuration: {"Detection Configuration":"Base"}. This record does not assert independent reproduction. ### InternVL3-38B Mirage Rate: 27.1. Result ID: 6a1e900afbc373592c1894c6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Detection Configuration":"Base"}. This record does not assert independent reproduction. ### Qwen2.5-VL-32B Mirage Rate: 57.4. Result ID: 6a1e900afbc373592c1894ae. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Detection Configuration":"Base"}. This record does not assert independent reproduction. ### BLIP2-2.7B Mirage Rate: 58. Result ID: 6a1e900afbc373592c1894cf. [Reporting paper]() Recorded evaluation configuration: {"Detection Configuration":"Base"}. This record does not assert independent reproduction. ### Gemma3-4B Mirage Rate: 58.8. Result ID: 6a1e900afbc373592c1894bd. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Detection Configuration":"Base"}. This record does not assert independent reproduction. ### Phi-3.5-Vision Mirage Rate: 61.9. Result ID: 6a1e900afbc373592c1894cc. [Reporting paper]() Recorded evaluation configuration: {"Detection Configuration":"Base"}. This record does not assert independent reproduction. ### Qwen2.5-VL-72B Mirage Rate: 63.6. Result ID: 6a1e900afbc373592c1894b1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Detection Configuration":"Base"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Mirage Rate: 63.6. Result ID: 6a1e900afbc373592c1894ba. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Detection Configuration":"Base"}. This record does not assert independent reproduction. ### LLaVA-Next-110B Mirage Rate: 63.6. Result ID: 6a1e900afbc373592c1894c3. [Reporting paper]() Recorded evaluation configuration: {"Detection Configuration":"Base"}. This record does not assert independent reproduction. ### MiniCPM-V-2.6 Mirage Rate: 66.6. Result ID: 6a1e900afbc373592c1894c9. [Reporting paper]() Recorded evaluation configuration: {"Detection Configuration":"Base"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.