# Visual Question Answering on VQA v2 (Probing Refusal Metrics) Source: [Visual Question Answering on VQA v2 (Probing Refusal Metrics)]() Collection index: [llms.txt]() Data updated: 2026-04-13T08:39:55.436Z Evaluation of model performance on the VQA v2 dataset for visual question answering, focusing specifically on refusal rates across different types of probing questions (Exact, Similar, General). 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 | Visual Question Answering | | Dataset (including any stated split) | VQA v2 | | Metric | Refusal Rate (Exact) | | Unit | PERCENT | | Direction | Lower is better | | Results loaded | 5 of 5 | | 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 Rate (Exact)](): Lower is better; unit: PERCENT. - [Refusal Rate (Similar)](): Lower is better; unit: PERCENT. - [Refusal Rate (General)](): Lower is better; unit: PERCENT. ## Recorded results | Method | Refusal Rate (Exact) | Result date | Report paper | | --- | --- | --- | --- | | No Perturbation | 0 | Not recorded | 69dc78627d0172140ec7891f | | Qi et al. | 2 | Not recorded | 69dc78627d0172140ec7891f | | Bagdasaryan et al. | 65 | Not recorded | 69dc78627d0172140ec7891f | | ImageProtector+PGD | 94 | Not recorded | 69dc78627d0172140ec7891f | | ImageProtector | 94 | Not recorded | 69dc78627d0172140ec7891f | ### No Perturbation Refusal Rate (Exact): 0. Result ID: 69dcaac147f34ce924c3a181. [Reporting paper]() Recorded evaluation configuration: {"Target MLLM":"LLaVA-1.5"}. This record does not assert independent reproduction. ### Qi et al. Refusal Rate (Exact): 2. Result ID: 69dcaac147f34ce924c3a182. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target MLLM":"LLaVA-1.5"}. This record does not assert independent reproduction. ### Bagdasaryan et al. Refusal Rate (Exact): 65. Result ID: 69dcaac147f34ce924c3a183. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target MLLM":"LLaVA-1.5"}. This record does not assert independent reproduction. ### ImageProtector+PGD Refusal Rate (Exact): 94. Result ID: 69dcaac147f34ce924c3a184. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target MLLM":"LLaVA-1.5"}. This record does not assert independent reproduction. ### ImageProtector Refusal Rate (Exact): 94. Result ID: 69dcaac147f34ce924c3a185. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target MLLM":"LLaVA-1.5"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.