# Text-based Visual Question Answering on TextVQA (Adversarial & Robustness Metrics) Source: [Text-based Visual Question Answering on TextVQA (Adversarial & Robustness Metrics)]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:52:38.511Z Evaluation of Text-based Visual Question Answering performance on the TextVQA dataset using a specific set of metrics covering clean, adversarial, and robustness measures (M(cl,nc), M(adv,c), UPR, CAE, CSG). 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 | Text-based Visual Question Answering | | Dataset (including any stated split) | TextVQA | | Metric | Metric (Clean, No Context) | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 9 of 9 | | 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 - [Metric (Clean, No Context)](): Higher is better; unit: PERCENT. - [Metric (Clean, Context)](): Higher is better; unit: PERCENT. - [Metric (Adversarial, No Context)](): Higher is better; unit: PERCENT. - [Metric (Adversarial, Context)](): Higher is better; unit: PERCENT. - [UPR](): Higher is better; unit: PERCENT. - [CAE](): Higher is better; unit: PERCENT. - [CSG](): Higher is better; unit: SCALAR. ## Recorded results | Method | Metric (Clean, No Context) | Result date | Report paper | | --- | --- | --- | --- | | Vanilla Attack | 91.35 | Not recorded | 69705d5de25ad2fa9d1ece20 | | Random Attack | 91.35 | Not recorded | 69705d5de25ad2fa9d1ece20 | | CAA | 91.35 | Not recorded | 69705d5de25ad2fa9d1ece20 | | Vanilla Attack | 88.17 | Not recorded | 69705d5de25ad2fa9d1ece20 | | Random Attack | 88.17 | Not recorded | 69705d5de25ad2fa9d1ece20 | | CAA | 88.17 | Not recorded | 69705d5de25ad2fa9d1ece20 | | Vanilla Attack | 63.18 | Not recorded | 69705d5de25ad2fa9d1ece20 | | Random Attack | 63.18 | Not recorded | 69705d5de25ad2fa9d1ece20 | | CAA | 63.18 | Not recorded | 69705d5de25ad2fa9d1ece20 | ### Vanilla Attack Metric (Clean, No Context): 91.35. Result ID: 6996ec5f5698665ec438e3b4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen-VL","Attack Type":"Vanilla","Compression":"PDrop"}. This record does not assert independent reproduction. ### Random Attack Metric (Clean, No Context): 91.35. Result ID: 6996ec5f5698665ec438e3b5. [Reporting paper]() Recorded evaluation configuration: {"Model":"Qwen-VL","Attack Type":"Random","Compression":"PDrop"}. This record does not assert independent reproduction. ### CAA Metric (Clean, No Context): 91.35. Result ID: 6996ec5f5698665ec438e3b6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen-VL","Attack Type":"Compression-Aware","Compression":"PDrop"}. This record does not assert independent reproduction. ### Vanilla Attack Metric (Clean, No Context): 88.17. Result ID: 6996ec5f5698665ec438e3ae. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLaVA","Attack Type":"Vanilla","Compression":"PDrop"}. This record does not assert independent reproduction. ### Random Attack Metric (Clean, No Context): 88.17. Result ID: 6996ec5f5698665ec438e3af. [Reporting paper]() Recorded evaluation configuration: {"Model":"LLaVA","Attack Type":"Random","Compression":"PDrop"}. This record does not assert independent reproduction. ### CAA Metric (Clean, No Context): 88.17. Result ID: 6996ec5f5698665ec438e3b0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLaVA","Attack Type":"Compression-Aware","Compression":"PDrop"}. This record does not assert independent reproduction. ### Vanilla Attack Metric (Clean, No Context): 63.18. Result ID: 6996ec5f5698665ec438e3b1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLaVA-NEXT","Attack Type":"Vanilla","Compression":"PDrop"}. This record does not assert independent reproduction. ### Random Attack Metric (Clean, No Context): 63.18. Result ID: 6996ec5f5698665ec438e3b2. [Reporting paper]() Recorded evaluation configuration: {"Model":"LLaVA-NEXT","Attack Type":"Random","Compression":"PDrop"}. This record does not assert independent reproduction. ### CAA Metric (Clean, No Context): 63.18. Result ID: 6996ec5f5698665ec438e3b3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLaVA-NEXT","Attack Type":"Compression-Aware","Compression":"PDrop"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.