Research map / Vision-language and multimodal / Cross-modal learning
Two-Stage Multi-Modal Fusion with Adaptive Alignment for Action Quality Assessment
Kanglei Zhou, Ruizhi Cai, Xinning Wang, Yijian Zheng, Liyuan Wang, Jianguo Li, Xiaohui Liang
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 |
|---|---|---|---|
| use DualAlign two-stage multi-modal alignment instead of best unimodal baseline | DualAlign · MM-JDM | SRCC x100: 64.28 -> 85.44 | Improved |
| use DualAlign instead of strongest multi-modal baseline | DualAlign · MM-JDM | SRCC x100: 60.30 -> 85.44 | Improved |
| use DualAlign instead of DAE | DualAlign · MM-JDM | rMSE: 3.036 -> 2.363 | Improved |
| use DualAlign instead of best unimodal baseline | DualAlign · MM-JDM | accuracy (%): 81.64 -> 87.41 | Improved |
| use DualAlign instead of MLAVL on RG | DualAlign · RG | SRCC: 0.849 -> 0.878 | Improved |
| use DualAlign instead of MLAVL on Fis-V | DualAlign · Fis-V | SRCC: 0.823 -> 0.872 | Improved |
| use DualAlign instead of MLAVL on Fis-V | DualAlign · Fis-V | MSE: 13.31 -> 11.66 | Improved |
| use DualAlign instead of unimodal CoRe on Action 10 | DualAlign · MM-JDM Action 10 | SRCC x100: 82.26 -> 70.56 | Hurt |
| remove the RGB video stream (ablation) | DualAlign · MM-JDM | SRCC: dropping by 45% | Improved |
| remove the RGB video stream (ablation) | DualAlign · MM-JDM | rMSE: increasing by 187% | Improved |
| remove the RGB video stream (ablation) | DualAlign · MM-JDM | accuracy: decreasing by 23% | Improved |
| use zero-shot GPT-4o or Gemini-3 Pro instead of trained DualAlign | mLLMs · MM-JDM | accuracy (%): 20.29% and 36.87% vs 87.41% | Hurt |
Closest papers by meaning
- BriMA: Bridged Modality Adaptation for Multi-Modal Continual Action Quality Assessment (CVPR 2026 · similarity 0.90)
- LucidAction: A Hierarchical and Multi-model Dataset for Comprehensive Action Quality Assessment (NeurIPS 2024 · similarity 0.87)
- MotionBench: Benchmarking and Improving Fine-grained Video Motion Understanding for Vision Language Models (CVPR 2025 · similarity 0.85)
- AlignMamba: Enhancing Multimodal Mamba with Local and Global Cross-modal Alignment (CVPR 2025 · similarity 0.85)
- Video-MME: The First-Ever Comprehensive Evaluation Benchmark of Multi-modal LLMs in Video Analysis (CVPR 2025 · similarity 0.85)
- When to Align, When to Predict: A Phase Diagram for Multimodal Learning (NeurIPS 2026 · similarity 0.85)
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.
