Research map / Vision-language and multimodal / Visual question answering
CLIP-Powered TASS: Target-Aware Single-Stream Network for Audio-Visual Question Answering
Yuanyuan Jiang, Jianqin Yin
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 target-aware process to spatial grounding (ablation) | TASS · MUSIC-AVQA (CLIP features) | accuracy, all questions (%): 74.27 -> 74.98 | Improved |
| add audio-visual match loss L_s (ablation) | TASS · MUSIC-AVQA (CLIP features) | accuracy, all questions (%): 72.40 -> 74.98 | Improved |
| add cross-modal synchrony loss (ablation) | TASS · MUSIC-AVQA (CLIP features) | accuracy, all questions (%): 74.22 -> 74.98 | Improved |
| add TSG+ module with L_s (ablation) | TASS · MUSIC-AVQA (CLIP features) | accuracy, all questions (%): 73.81 -> 74.98 | Improved |
| add JTG single-stream module with L_cms (ablation) | TASS · MUSIC-AVQA (CLIP features) | accuracy, all questions (%): 73.64 -> 74.98 | Improved |
| add TSG+ module without its match loss L_s (CLIP unconstrained) (ablation) | TASS · MUSIC-AVQA (CLIP features) | accuracy, all questions (%): 73.81 -> 72.40 | Hurt |
| add temporal pooling preprocessing (ablation) | TASS (ResNet-18 + GloVe) · MUSIC-AVQA (ResNet-18 + GloVe features) | accuracy, all questions (%): 73.05 -> 73.63 | Improved |
| add temporal pooling preprocessing (ablation) | TASS · MUSIC-AVQA (CLIP features) | accuracy, all questions (%): 74.09 -> 74.98 | Improved |
| use additive target-aware attention instead of Hadamard product (ablation) | TSG+ · MUSIC-AVQA (CLIP features) | accuracy, all questions (%): 74.63 -> 74.98 | Improved |
| use threshold tau 0.025 instead of 0.030 (ablation) | TSG+ · MUSIC-AVQA (CLIP features) | accuracy, all questions (%): 74.30 -> 74.98 | Improved |
| use interleaved visual-audio sequence instead of concatenation (ablation) | JTG · MUSIC-AVQA (CLIP features) | accuracy, all questions (%): 74.38 -> 74.98 | Improved |
| use single-stream instead of dual-stream with CMS (ablation) | TASS · MUSIC-AVQA (CLIP features) | accuracy, all questions (%): 73.68 -> 74.98 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- COCA — COCA: COllaborative CAusal Regularization for Audio-Visual Question Answering (AAAI 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Audio-Visual Adaptive Fusion Network for Question Answering Based on Contrastive Learning (AAAI 2025 · similarity 0.90)
- AVQACL: A Novel Benchmark for Audio-Visual Question Answering Continual Learning (CVPR 2025 · similarity 0.90)
- Object-Aware Adaptive-Positivity Learning for Audio-Visual Question Answering (AAAI 2024 · similarity 0.89)
- Look, Listen, and Answer: Overcoming Biases for Audio-Visual Question Answering (NeurIPS 2024 · similarity 0.89)
- OmniVideoBench: Towards Audio-Visual Understanding Evaluation for Omni MLLMs (ICLR 2026 · similarity 0.89)
- 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.
