# Audio-visual QA on MUSIC-AVQA Source: [Audio-visual QA on MUSIC-AVQA]() Collection index: [llms.txt]() Data updated: 2026-06-02T08:19:08.452Z Evaluation of model performance on the MUSIC-AVQA dataset for Audio-visual Question Answering, reporting separate accuracies for audio-only, visual-only, and combined inputs. 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 QA | | Dataset (including any stated split) | MUSIC-AVQA | | Metric | Overall Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 11 of 11 | | 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 - [Overall Accuracy](): Higher is better; unit: PERCENT. - [Audio Accuracy](): Higher is better; unit: PERCENT. - [Visual Accuracy](): Higher is better; unit: PERCENT. - [Audio-Visual Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | Overall Accuracy | Result date | Report paper | | --- | --- | --- | --- | | CAT-7B-FT | 84.3 | Not recorded | 6a1e63bfdbd7150506f85346 | | V-LynX-7B | 83 | Not recorded | 6a1e63bfdbd7150506f85346 | | PAVE-7B | 82.3 | Not recorded | 6a1e63bfdbd7150506f85346 | | V-LynX-0.5B | 81.1 | Not recorded | 6a1e63bfdbd7150506f85346 | | VAST | 80.7 | Not recorded | 6a1e63bfdbd7150506f85346 | | PAVE-0.5B | 78.8 | Not recorded | 6a1e63bfdbd7150506f85346 | | LLaVA-OV-7B-FT | 77.4 | Not recorded | 6a1e63bfdbd7150506f85346 | | LLaVA-OV-0.5B-FT | 67.6 | Not recorded | 6a1e63bfdbd7150506f85346 | | LLaVA-OV-7B | 60.4 | Not recorded | 6a1e63bfdbd7150506f85346 | | LLaVA-OV-0.5B | 52.8 | Not recorded | 6a1e63bfdbd7150506f85346 | | CAT-7B | 48.6 | Not recorded | 6a1e63bfdbd7150506f85346 | ### CAT-7B-FT Overall Accuracy: 84.3. Result ID: 6a1e91f0fbc373592c18a4ad. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Evaluation Protocol":"Task-specific","Model Scale":"≥ 7B"}. This record does not assert independent reproduction. ### V-LynX-7B Overall Accuracy: 83. Result ID: 6a1e91f0fbc373592c18a4b0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Protocol":"Task-specific","Model Scale":"≥ 7B","∆Params.":"195.0M"}. This record does not assert independent reproduction. ### PAVE-7B Overall Accuracy: 82.3. Result ID: 6a1e91f0fbc373592c18a4af. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Evaluation Protocol":"Task-specific","Model Scale":"≥ 7B","∆Params.":"256.7M"}. This record does not assert independent reproduction. ### V-LynX-0.5B Overall Accuracy: 81.1. Result ID: 6a1e91f0fbc373592c18a4ac. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Protocol":"Task-specific","Model Scale":"\< 7B","∆Params.":"68.7M"}. This record does not assert independent reproduction. ### VAST Overall Accuracy: 80.7. Result ID: 6a1e91f0fbc373592c18a4a9. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Evaluation Protocol":"Task-specific","Model Scale":"\< 7B"}. This record does not assert independent reproduction. ### PAVE-0.5B Overall Accuracy: 78.8. Result ID: 6a1e91f0fbc373592c18a4ab. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Evaluation Protocol":"Task-specific","Model Scale":"\< 7B","∆Params.":"127.6M"}. This record does not assert independent reproduction. ### LLaVA-OV-7B-FT Overall Accuracy: 77.4. Result ID: 6a1e91f0fbc373592c18a4ae. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Evaluation Protocol":"Task-specific","Model Scale":"≥ 7B","∆Params.":"161.5M"}. This record does not assert independent reproduction. ### LLaVA-OV-0.5B-FT Overall Accuracy: 67.6. Result ID: 6a1e91f0fbc373592c18a4aa. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Evaluation Protocol":"Task-specific","Model Scale":"\< 7B","∆Params.":"35.2M"}. This record does not assert independent reproduction. ### LLaVA-OV-7B Overall Accuracy: 60.4. Result ID: 6a1e91f0fbc373592c18a4a8. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Evaluation Protocol":"Zero-shot"}. This record does not assert independent reproduction. ### LLaVA-OV-0.5B Overall Accuracy: 52.8. Result ID: 6a1e91f0fbc373592c18a4a7. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Evaluation Protocol":"Zero-shot"}. This record does not assert independent reproduction. ### CAT-7B Overall Accuracy: 48.6. Result ID: 6a1e91f0fbc373592c18a4a6. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Evaluation Protocol":"Zero-shot"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.