Research map / Vision-language and multimodal / Cross-modal learning
Active Exploration of Multimodal Complementarity for Few-Shot Action Recognition
Yuyang Wanyan, Xiaoshan Yang, Chaofan Chen, Changsheng Xu
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 AMFAR's active modality selection, mutual distillation, and adaptive inference instead of the strongest listed baseline | few-shot action recognition · Kinetics 5-way 1-shot | accuracy: 74.6 -> 80.1 | Improved |
| Use AMFAR's active modality selection, mutual distillation, and adaptive inference instead of the strongest listed baseline | few-shot action recognition · Kinetics 5-way 5-shot | accuracy: 87.4 -> 92.6 | Improved |
| Use AMFAR's active modality selection, mutual distillation, and adaptive inference instead of the strongest listed baseline | few-shot action recognition · SSv2 5-way 1-shot | accuracy: 54.3 -> 61.7 | Improved |
| Use AMFAR's active modality selection, mutual distillation, and adaptive inference instead of the strongest listed baseline | few-shot action recognition · SSv2 5-way 5-shot | accuracy: 70.4 -> 79.5 | Improved |
| Use AMFAR's active modality selection, mutual distillation, and adaptive inference instead of the strongest listed baseline | few-shot action recognition · HMDB51 5-way 1-shot | accuracy: 60.3 -> 73.9 | Improved |
| Use AMFAR's active modality selection, mutual distillation, and adaptive inference instead of the strongest listed baseline | few-shot action recognition · HMDB51 5-way 5-shot | accuracy: 81.3 -> 87.8 | Improved |
| Use AMFAR's active modality selection, mutual distillation, and adaptive inference instead of the strongest listed baseline | few-shot action recognition · UCF101 5-way 1-shot | accuracy: 87.4 -> 91.2 | Improved |
| Use AMFAR's active modality selection, mutual distillation, and adaptive inference instead of the strongest listed baseline | few-shot action recognition · UCF101 5-way 5-shot | accuracy: 98.4 -> 99.0 | Improved |
| Add Active Sample Selection to group queries by the more reliable modality (ablation) | AMFAR · Kinetics 5-way 1-shot | accuracy: 72.9 -> 80.1 | Improved |
| Add Active Mutual Distillation between modalities (ablation) | AMFAR · SSv2 5-way 1-shot | accuracy: 58.8 -> 61.7 | Improved |
| Add Adaptive Multimodal Inference with reliability-weighted fusion (ablation) | AMFAR · SSv2 5-way 5-shot | accuracy: 78.2 -> 79.5 | Improved |
| Use absolute certainty to estimate modality reliability (ablation) | ASS in AMFAR · Kinetics 5-way 1-shot | accuracy: 72.9 -> 80.1 | Improved |
Closest papers by meaning
- Knowledge is Power: Advancing Few-shot Action Recognition with Multimodal Semantics from MLLMs (IJCV 2026 · similarity 0.90)
- Beyond Label Semantics: Language-Guided Action Anatomy for Few-shot Action Recognition (ICCV 2025 · similarity 0.87)
- Multimodality Helps Unimodality: Cross-Modal Few-Shot Learning With Multimodal Models (CVPR 2023 · similarity 0.87)
- Learning Causal Domain-Invariant Temporal Dynamics for Few-Shot Action Recognition (ICML 2024 · similarity 0.86)
- Context-Aware Multimodal Pretraining (CVPR 2025 · similarity 0.85)
- A Multimodal, Multi-Task Adapting Framework for Video Action Recognition (AAAI 2024 · 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.
