| # Audio Visual Learning: Research Notes | |
| ## Status | |
| Working note / experiment plan. No completed benchmark results are claimed here. | |
| ## 1. Scope and motivation | |
| These notes organize a possible evaluation of temporal correspondence across video, language, and audio. The central question is whether the proposed change improves the target behavior under a matched training and evaluation budget. The note deliberately separates hypotheses from observations so that future results can be added without rewriting the rationale. | |
| ## 2. Context | |
| Research on audio visual learning often mixes improvements from architecture, data scale, preprocessing, and compute. A useful comparison therefore needs controlled baselines and explicit reporting of resource use. For this topic, the main confound is that clip sampling and timestamp noise can hide failures on long-range events. | |
| ## 3. Working hypothesis | |
| A focused change to the representation or interaction mechanism may improve Recall@K without increasing deployment cost disproportionately. The hypothesis should be rejected if gains disappear after matching parameter count, data exposure, or tuning budget. | |
| ## 4. Proposed approach | |
| The first implementation should keep modality-specific preprocessing simple, project inputs into a shared representation space, and isolate the new component behind a small interface. Baselines should include a comparable model without the component and a stronger off-the-shelf reference. Any optimization should be applied to all systems, not only the proposed one. | |
| ## 5. Evaluation plan | |
| | Dataset | Role | Primary measure | | |
| |---|---|---| | |
| | MSR-VTT | primary evaluation | Recall@K | | |
| | ActivityNet Captions | transfer / robustness | CIDEr | | |
| | VGGSound | transfer / robustness | mean average precision | | |
| Planned comparisons include a matched-capacity baseline, an ablation that removes the proposed component, and an out-of-domain transfer check. Default training values for the first controlled run are learning rate `0.0002`, batch size `48`, and `5` independent seeds. These are planning values, not claims about a finished experiment. | |
| ## 6. Reproducibility checklist | |
| - Fix preprocessing before tuning. | |
| - Report mean and standard deviation across seeds. | |
| - Keep a held-out error-analysis split. | |
| - Record wall-clock time and peak memory. | |
| ## 7. Failure modes and responsible use | |
| The analysis should report subgroup and category-level failures instead of relying only on a single aggregate score. Particular attention is needed because clip sampling and timestamp noise can hide failures on long-range events. No production use is recommended without task-specific validation, data review, and an assessment of privacy and bias. | |
| ## 8. Open questions | |
| - Where does the method fail on compositional or out-of-domain examples? | |
| - Which gain survives when the compute budget is matched? | |
| - Does the proposed component improve calibration as well as the primary metric? | |
| ## References | |
| [1] Xu et al., MSR-VTT, 2016. | |
| [2] Krishna et al., ActivityNet Captions, 2017. | |
| [3] Chen et al., VGGSound, 2020. | |