# Audio-Visual Question Answering on AVQA (subset 2000 samples) Source: [Audio-Visual Question Answering on AVQA (subset 2000 samples)]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:24:04.247Z Evaluation of Audio-Visual Question Answering performance using the 2000-sample subset of the AVQA dataset, reporting ASR metrics. 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 | Audio-Visual Question Answering | | Dataset (including any stated split) | AVQA (subset 2000 samples) | | Metric | ASR Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 7 of 7 | | 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 - [ASR Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | ASR Accuracy | Result date | Report paper | | --- | --- | --- | --- | | Combined Loss | 96.03 | Not recorded | 6976f519a2b28e761d87c22c | | Encoder-Based Cosine Similarity Loss | 89.12 | Not recorded | 6976f519a2b28e761d87c22c | | Audio Attention Amplification Loss | 56.21 | Not recorded | 6976f519a2b28e761d87c22c | | Vision Attention Suppression Loss | 18.72 | Not recorded | 6976f519a2b28e761d87c22c | | Attention Randomization Loss | 17.24 | Not recorded | 6976f519a2b28e761d87c22c | | Hidden-State Similarity Loss | 15.13 | Not recorded | 6976f519a2b28e761d87c22c | | Negative Language Modeling Loss | 10.27 | Not recorded | 6976f519a2b28e761d87c22c | ### Combined Loss ASR Accuracy: 96.03. Result ID: 699a052b5698665ec4878b89. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Attack Objective":"L(combined)","Source Model":"VideoLLAMA2","Training Dataset":"AVQA","Training Epochs":"150"}. This record does not assert independent reproduction. ### Encoder-Based Cosine Similarity Loss ASR Accuracy: 89.12. Result ID: 699a052b5698665ec4878b84. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Attack Objective":"L(cos)","Source Model":"VideoLLAMA2","Training Dataset":"AVQA","Training Epochs":"150"}. This record does not assert independent reproduction. ### Audio Attention Amplification Loss ASR Accuracy: 56.21. Result ID: 699a052b5698665ec4878b86. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Attack Objective":"L(audioatt)","Source Model":"VideoLLAMA2","Training Dataset":"AVQA","Training Epochs":"150"}. This record does not assert independent reproduction. ### Vision Attention Suppression Loss ASR Accuracy: 18.72. Result ID: 699a052b5698665ec4878b85. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Attack Objective":"L(visionatt)","Source Model":"VideoLLAMA2","Training Dataset":"AVQA","Training Epochs":"150"}. This record does not assert independent reproduction. ### Attention Randomization Loss ASR Accuracy: 17.24. Result ID: 699a052b5698665ec4878b87. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Attack Objective":"L(randatt)","Source Model":"VideoLLAMA2","Training Dataset":"AVQA","Training Epochs":"150"}. This record does not assert independent reproduction. ### Hidden-State Similarity Loss ASR Accuracy: 15.13. Result ID: 699a052b5698665ec4878b88. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Attack Objective":"L(hidden-cos)","Source Model":"VideoLLAMA2","Training Dataset":"AVQA","Training Epochs":"150"}. This record does not assert independent reproduction. ### Negative Language Modeling Loss ASR Accuracy: 10.27. Result ID: 699a052b5698665ec4878b83. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Attack Objective":"L\_negLM","Source Model":"VideoLLAMA2","Training Dataset":"AVQA","Training Epochs":"150"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.