Research map / Vision-language and multimodal / Visual question answering
Pano-AVQA: Grounded Audio-Visual Question Answering on 360deg Videos
Heeseung Yun, Youngjae Yu, Wonsuk Yang, Kangil Lee, Gunhee Kim
Read the paperCheck your idea against 137,217 papers
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 |
|---|---|---|---|
| add unimodal encoders before multimodal fusion (ablation) | LAViT · Pano-AVQA test | accuracy (all): 44.14 -> 50.49 | Improved |
| add visual input to audio+language (ablation) | LAViT · Pano-AVQA test | accuracy (all): 47.99 -> 50.49 | Improved |
| add audio input to visual+language (ablation) | LAViT · Pano-AVQA test | accuracy (all): 48.73 -> 50.49 | Improved |
| use both equirectangular and NFoV views instead of single NFoV (ablation) | LAViT · Pano-AVQA test | accuracy (all): 48.50 -> 50.49 | Improved |
| add NFoV views to equirectangular only (ablation) | LAViT · Pano-AVQA test | accuracy (all): 49.18 -> 50.49 | Improved |
| add equirectangular view to dense NFoV (ablation) | LAViT · Pano-AVQA test | accuracy (all): 49.79 -> 50.49 | Improved |
| use quaternion instead of Cartesian spatial embedding (ablation) | LAViT · Pano-AVQA test | accuracy (all): 49.89 -> 50.49 | Improved |
| use quaternion instead of spherical coordinates embedding (ablation) | LAViT · Pano-AVQA test | accuracy (all): 50.21 -> 50.49 | No clear effect |
| use quaternion instead of unit sphere embedding (ablation) | LAViT · Pano-AVQA test | accuracy (all): 49.83 -> 50.49 | Improved |
| use early multimodal transformer fusion instead of late fusion (LAViT vs AVSD) (ablation) | LAViT · Pano-AVQA test | accuracy (all): 24.60 -> 50.49 | Improved |
| use LAViT with audio instead of LXMERT (ablation) | LAViT vs LXMERT · Pano-AVQA test | accuracy (all): 48.48 -> 50.49 | Improved |
Closest papers by meaning
- AVQACL: A Novel Benchmark for Audio-Visual Question Answering Continual Learning (CVPR 2025 · similarity 0.88)
- MAGNET: A Multi-agent Framework for Finding Audio-Visual Needles by Reasoning over Multi-Video Haystacks (NeurIPS 2025 · similarity 0.88)
- Bridging the Gap between 2D and 3D Visual Question Answering: A Fusion Approach for 3D VQA (AAAI 2024 · similarity 0.87)
- Object-Aware Adaptive-Positivity Learning for Audio-Visual Question Answering (AAAI 2024 · similarity 0.87)
- Audio-Visual Adaptive Fusion Network for Question Answering Based on Contrastive Learning (AAAI 2025 · similarity 0.87)
- Question-Aware Gaussian Experts for Audio-Visual Question Answering (CVPR 2025 · similarity 0.87)
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.
