# Audio Visual Question Answering on Music-AVQA (Robustness Evaluation) Source: [Audio Visual Question Answering on Music-AVQA (Robustness Evaluation)]() Collection index: [llms.txt]() Data updated: 2026-04-07T08:51:18.804Z Evaluation of model performance on the Music-AVQA dataset for Audio Visual Question Answering, focusing on robustness by reporting clean accuracy, adversarial attack accuracy, and the resulting accuracy drop. 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) | Music-AVQA | | Metric | Music-AVQA Clean Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 25 of 25 | | 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 - [Music-AVQA Clean Accuracy](): Higher is better; unit: PERCENT. - [Music-AVQA Attack Accuracy](): Higher is better; unit: PERCENT. - [Music-AVQA Accuracy Drop](): Lower is better; unit: PERCENT. - [Music-AVQA ASR Clean](): Lower is better; unit: SCALAR. - [Music-AVQA ASR Attack](): Lower is better; unit: SCALAR. ## Recorded results | Method | Music-AVQA Clean Accuracy | Result date | Report paper | | --- | --- | --- | --- | | Qwen3-Omni-30B | 85.15 | Not recorded | 69d48f7b7d0172140ec742a9 | | Qwen2.5-Omni-7B | 82.99 | Not recorded | 69d48f7b7d0172140ec742a9 | | Gemini-3.1-Flash-Lite-preview | 81.32 | Not recorded | 69d48f7b7d0172140ec742a9 | | Negative Language Modeling Loss | 80.7 | Not recorded | 6976f519a2b28e761d87c22c | | Encoder-Based Cosine Similarity Loss | 80.7 | Not recorded | 6976f519a2b28e761d87c22c | | Vision Attention Suppression Loss | 80.7 | Not recorded | 6976f519a2b28e761d87c22c | | Audio Attention Amplification Loss | 80.7 | Not recorded | 6976f519a2b28e761d87c22c | | Attention Randomization Loss | 80.7 | Not recorded | 6976f519a2b28e761d87c22c | | Hidden-State Similarity Loss | 80.7 | Not recorded | 6976f519a2b28e761d87c22c | | Combined Loss (SOUNDBREAK) | 80.7 | Not recorded | 6976f519a2b28e761d87c22c | | Gemini-2.5-Flash-Lite | 80.68 | Not recorded | 69d48f7b7d0172140ec742a9 | | Gemini-3.1-Flash-Lite-preview | 71.97 | Not recorded | 69d48f7b7d0172140ec742a9 | | Gemini-2.5-Flash-Lite | 68.99 | Not recorded | 69d48f7b7d0172140ec742a9 | | Qwen2.5-Omni-7B | 66.94 | Not recorded | 69d48f7b7d0172140ec742a9 | | PandaGPT | 64.41 | Not recorded | 69d48f7b7d0172140ec742a9 | | Gemini-3.1-Flash-Lite-preview | 62.63 | Not recorded | 69d48f7b7d0172140ec742a9 | | Qwen3-Omni-30B | 61.54 | Not recorded | 69d48f7b7d0172140ec742a9 | | Gemini-2.5-Flash-Lite | 60.15 | Not recorded | 69d48f7b7d0172140ec742a9 | | Qwen2.5-Omni-7B | 57.01 | Not recorded | 69d48f7b7d0172140ec742a9 | | Qwen3-Omni-30B | 56.57 | Not recorded | 69d48f7b7d0172140ec742a9 | | ChatBridge | 51.38 | Not recorded | 69d48f7b7d0172140ec742a9 | | ChatBridge | 39.99 | Not recorded | 69d48f7b7d0172140ec742a9 | | ChatBridge | 37.64 | Not recorded | 69d48f7b7d0172140ec742a9 | | PandaGPT | 35.98 | Not recorded | 69d48f7b7d0172140ec742a9 | | PandaGPT | 34.93 | Not recorded | 69d48f7b7d0172140ec742a9 | ### Qwen3-Omni-30B Music-AVQA Clean Accuracy: 85.15. Result ID: 69d4c56847f34ce924bcab94. [Reporting paper]() Recorded evaluation configuration: {"Model":"Qwen3-Omni-30B"}. This record does not assert independent reproduction. ### Qwen2.5-Omni-7B Music-AVQA Clean Accuracy: 82.99. Result ID: 69d4c56847f34ce924bcab93. [Reporting paper]() Recorded evaluation configuration: {"Model":"Qwen2.5-Omni-7B"}. This record does not assert independent reproduction. ### Gemini-3.1-Flash-Lite-preview Music-AVQA Clean Accuracy: 81.32. Result ID: 69d4c56847f34ce924bcab98. [Reporting paper]() Recorded evaluation configuration: {"Model":"Gemini-3.1-Flash-Lite-preview"}. This record does not assert independent reproduction. ### Negative Language Modeling Loss Music-AVQA Clean Accuracy: 80.7. Result ID: 699a052b5698665ec4878b7c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Objective":"L\_negLM"}. This record does not assert independent reproduction. ### Encoder-Based Cosine Similarity Loss Music-AVQA Clean Accuracy: 80.7. Result ID: 699a052b5698665ec4878b7d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Objective":"L^(cos)"}. This record does not assert independent reproduction. ### Vision Attention Suppression Loss Music-AVQA Clean Accuracy: 80.7. Result ID: 699a052b5698665ec4878b7e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Objective":"L^(visionatt)"}. This record does not assert independent reproduction. ### Audio Attention Amplification Loss Music-AVQA Clean Accuracy: 80.7. Result ID: 699a052b5698665ec4878b7f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Objective":"L^(audioatt)"}. This record does not assert independent reproduction. ### Attention Randomization Loss Music-AVQA Clean Accuracy: 80.7. Result ID: 699a052b5698665ec4878b80. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Objective":"L^(randatt)"}. This record does not assert independent reproduction. ### Hidden-State Similarity Loss Music-AVQA Clean Accuracy: 80.7. Result ID: 699a052b5698665ec4878b81. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Objective":"L^(hidden-cos)"}. This record does not assert independent reproduction. ### Combined Loss (SOUNDBREAK) Music-AVQA Clean Accuracy: 80.7. Result ID: 699a052b5698665ec4878b82. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Objective":"L^(combined)"}. This record does not assert independent reproduction. ### Gemini-2.5-Flash-Lite Music-AVQA Clean Accuracy: 80.68. Result ID: 69d4c56847f34ce924bcab97. [Reporting paper]() Recorded evaluation configuration: {"Model":"Gemini-2.5-Flash-Lite"}. This record does not assert independent reproduction. ### Gemini-3.1-Flash-Lite-preview Music-AVQA Clean Accuracy: 71.97. Result ID: 69d4c56847f34ce924bcab92. [Reporting paper]() Recorded evaluation configuration: {"Model":"Gemini-3.1-Flash-Lite-preview"}. This record does not assert independent reproduction. ### Gemini-2.5-Flash-Lite Music-AVQA Clean Accuracy: 68.99. Result ID: 69d4c56847f34ce924bcab91. [Reporting paper]() Recorded evaluation configuration: {"Model":"Gemini-2.5-Flash-Lite"}. This record does not assert independent reproduction. ### Qwen2.5-Omni-7B Music-AVQA Clean Accuracy: 66.94. Result ID: 69d4c56847f34ce924bcab8d. [Reporting paper]() Recorded evaluation configuration: {"Model":"Qwen2.5-Omni-7B"}. This record does not assert independent reproduction. ### PandaGPT Music-AVQA Clean Accuracy: 64.41. Result ID: 69d4c56847f34ce924bcab95. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"PandaGPT"}. This record does not assert independent reproduction. ### Gemini-3.1-Flash-Lite-preview Music-AVQA Clean Accuracy: 62.63. Result ID: 69d4c56847f34ce924bcab9e. [Reporting paper]() Recorded evaluation configuration: {"Model":"Gemini-3.1-Flash-Lite-preview"}. This record does not assert independent reproduction. ### Qwen3-Omni-30B Music-AVQA Clean Accuracy: 61.54. Result ID: 69d4c56847f34ce924bcab8e. [Reporting paper]() Recorded evaluation configuration: {"Model":"Qwen3-Omni-30B"}. This record does not assert independent reproduction. ### Gemini-2.5-Flash-Lite Music-AVQA Clean Accuracy: 60.15. Result ID: 69d4c56847f34ce924bcab9d. [Reporting paper]() Recorded evaluation configuration: {"Model":"Gemini-2.5-Flash-Lite"}. This record does not assert independent reproduction. ### Qwen2.5-Omni-7B Music-AVQA Clean Accuracy: 57.01. Result ID: 69d4c56847f34ce924bcab99. [Reporting paper]() Recorded evaluation configuration: {"Model":"Qwen2.5-Omni-7B"}. This record does not assert independent reproduction. ### Qwen3-Omni-30B Music-AVQA Clean Accuracy: 56.57. Result ID: 69d4c56847f34ce924bcab9a. [Reporting paper]() Recorded evaluation configuration: {"Model":"Qwen3-Omni-30B"}. This record does not assert independent reproduction. ### ChatBridge Music-AVQA Clean Accuracy: 51.38. Result ID: 69d4c56847f34ce924bcab96. [Reporting paper]() Recorded evaluation configuration: {"Model":"ChatBridge"}. This record does not assert independent reproduction. ### ChatBridge Music-AVQA Clean Accuracy: 39.99. Result ID: 69d4c56847f34ce924bcab90. [Reporting paper]() Recorded evaluation configuration: {"Model":"ChatBridge"}. This record does not assert independent reproduction. ### ChatBridge Music-AVQA Clean Accuracy: 37.64. Result ID: 69d4c56847f34ce924bcab9c. [Reporting paper]() Recorded evaluation configuration: {"Model":"ChatBridge"}. This record does not assert independent reproduction. ### PandaGPT Music-AVQA Clean Accuracy: 35.98. Result ID: 69d4c56847f34ce924bcab8f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"PandaGPT"}. This record does not assert independent reproduction. ### PandaGPT Music-AVQA Clean Accuracy: 34.93. Result ID: 69d4c56847f34ce924bcab9b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"PandaGPT"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.