Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Multimodal Model-Agnostic Meta-Learning via Task-Aware Modulation
Risto Vuorio, Shao-Hua Sun, Hexiang Hu, Joseph J. Lim
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 |
|---|---|---|---|
| Condition MAML initialization on task-mode information using modulation | MAML · Multimodal few-shot image classification, 2-mode 5-way 1-shot | accuracy: 66.80% -> 69.93% | Improved |
| Condition MAML initialization on task-mode information using modulation | MAML · Multimodal few-shot image classification, 3-mode 5-way 5-shot | accuracy: 67.97% -> 70.15% | Improved |
| Condition MAML initialization on task-mode information using modulation | MAML · Multimodal few-shot image classification, 5-mode 5-way 5-shot | accuracy: 54.41% -> 60.83% | Improved |
| Use MMAML instead of MAML | MAML · Multimodal few-shot image classification, 2-mode 20-way 1-shot | accuracy: 44.69% -> 47.80% | Improved |
| Use MMAML instead of Multi-MAML with ground-truth mode labels | Multi-MAML · Multimodal few-shot image classification, 5-mode 5-way 5-shot | accuracy: 55.92% -> 60.83% | Improved |
| Use MMAML instead of MAML | MAML · Multimodal few-shot image classification, 3-mode 20-way 1-shot | accuracy: 28.22% -> 36.27% | Improved |
| Use MMAML instead of MAML | MAML · Multimodal few-shot image classification, 5-mode 20-way 1-shot | accuracy: 28.85% -> 33.97% | Improved |
| Use MMAML instead of Multi-MAML with ground-truth mode labels | Multi-MAML · Multimodal few-shot image classification, 2-mode 20-way 1-shot | accuracy: 53.15% -> 47.80% | Hurt |
| Use MMAML instead of Multi-MAML with ground-truth mode labels | Multi-MAML · Multimodal few-shot image classification, 3-mode 20-way 1-shot | accuracy: 39.77% -> 36.27% | Hurt |
| Add task-aware modulation to ProMP | ProMP · Point Mass 2D, 2 modes | cumulative reward per episode: -397 $\pm$ 20 -> -136 $\pm$ 8 | Improved |
| Add task-aware modulation to ProMP | ProMP · Point Mass 2D, 4 modes | cumulative reward per episode: -523 $\pm$ 51 -> -209 $\pm$ 32 | Improved |
| Add task-aware modulation to ProMP | ProMP · Reacher, 4 modes | cumulative reward per episode: -13.8 $\pm$ 2.5 -> -11.0 $\pm$ 0.8 | Improved |
Closest papers by meaning
- Learning to Learn Task-Adaptive Hyperparameters for Few-Shot Learning (TPAMI 2023 · similarity 0.90)
- Bridging Multi-Task Learning and Meta-Learning: Towards Efficient Training and Effective Adaptation (ICML 2021 · similarity 0.89)
- Learning to Forget for Meta-Learning via Task-and-Layer-Wise Attenuation (TPAMI 2021 · similarity 0.89)
- Revisit Multimodal Meta-Learning through the Lens of Multi-Task Learning (NeurIPS 2021 · similarity 0.89)
- Multimodal Representation Learning by Alternating Unimodal Adaptation (CVPR 2024 · similarity 0.88)
- Modality-Aware SAM: Sharpness-Aware-Minimization Driven Gradient Modulation for Harmonized Multimodal Learning (NeurIPS 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.
