Research map / Vision-language and multimodal / Visual question answering
Learning To Answer Questions in Dynamic Audio-Visual Scenarios
Guangyao Li, Yake Wei, Yapeng Tian, Chenliang Xu, Ji-Rong Wen, Di Hu
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use the proposed spatio-temporal grounding model instead of Pano-AVQA | AVQA on MUSIC-AVQA · MUSIC-AVQA test | accuracy (All Avg.): 68.93 -> 71.52 | Improved |
| Add audio input to question-only input (ablation) | AVQA model · MUSIC-AVQA test | accuracy (All): 54.09 -> 64.26 | Improved |
| Add visual input to question-only input (ablation) | AVQA model · MUSIC-AVQA test | accuracy (All): 54.09 -> 65.28 | Improved |
| Add both audio and visual inputs to question-only input (ablation) | AVQA model · MUSIC-AVQA test | accuracy (All): 54.09 -> 67.72 | Improved |
| Add temporal grounding (ablation) | AV+Q model · MUSIC-AVQA test | accuracy (All): 67.72 -> 70.27 | Improved |
| Add spatial grounding (ablation) | AV+Q+TG model · MUSIC-AVQA test | accuracy (All): 70.27 -> 71.52 | Improved |
| Shuffle video segments in time (ablation) | AV+Q+TG · MUSIC-AVQA temporal questions | accuracy: 65.17 -> 63.71 | Hurt |
| Add R(2+1)D motion features (ablation) | Our method · MUSIC-AVQA | accuracy: 71.53 -> 71.75 | Mixed |
| Add audio input to question-plus-video input (ablation) | Our method on TVQA · TVQA | accuracy: 42.01 -> 41.95 | Hurt |
| Add video and audio inputs to question-only input (ablation) | TVQA method · TVQA | accuracy: 43.50 -> 41.45 | Hurt |
Closest papers by meaning
- Look, Listen, and Answer: Overcoming Biases for Audio-Visual Question Answering (NeurIPS 2024 · similarity 0.93)
- Object-Aware Adaptive-Positivity Learning for Audio-Visual Question Answering (AAAI 2024 · similarity 0.93)
- Audio-Visual Adaptive Fusion Network for Question Answering Based on Contrastive Learning (AAAI 2025 · similarity 0.93)
- AVQACL: A Novel Benchmark for Audio-Visual Question Answering Continual Learning (CVPR 2025 · similarity 0.92)
- Question-Aware Gaussian Experts for Audio-Visual Question Answering (CVPR 2025 · similarity 0.91)
- JointAVBench: A Benchmark for Joint Audio-Visual Reasoning Evaluation (ICLR 2026 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
