Research map / Vision-language and multimodal / Cross-modal learning
Multimodal Generative Models for Scalable Weakly-Supervised Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the PoE inference network instead of the product-and-quotient variant (ablation) | MVAE · MNIST | log p(x1): -96.028 -> -90.619 | Helped |
| Treat each of 18 attributes as a separate modality (ablation) | MVAE · CelebA | log p(x1): -6236.923 -> -6236.109 | Helped |
| Increase the number of sampled ELBO terms k (ablation) | MVAE19 · CelebA | data log-likelihood | No clear effect |
| Increase the number of sampled ELBO terms k (ablation) | MVAE19 · CelebA | variance of importance weights | Helped |
| Use MVAE rather than JMVAE for weakly supervised prediction | MNIST · 1% paired examples | accuracy: 0.6116 -> 0.8838 | Helped |
| Use MVAE rather than JMVAE for weakly supervised prediction | FashionMNIST · 1% paired examples | accuracy: 0.6996 -> 0.8619 | Helped |
| Use MVAE rather than JMVAE for weakly supervised prediction | MultiMNIST · 1% paired examples | average digit accuracy: 0.1531 -> 0.2287 | Helped |
| Increase the fraction of aligned sentence pairs from 1% to 5% | MVAE · English-Vietnamese translation | test log p(x): -483.23 ± 5.81 -> -478.75 ± 3.00 | Helped |
| Increase the fraction of aligned sentence pairs from 5% to 100% | MVAE · English-Vietnamese translation | test log p(x): -478.75 ± 3.00 -> -478.12 ± 3.02 | No clear effect |
| Use MVAE rather than JMVAE for joint modeling | BinaryMNIST · BinaryMNIST | log p(x1,x2): -86.371 -> -86.255 | Helped |
| Use MVAE rather than JMVAE for joint modeling | MNIST · MNIST | log p(x1,x2): -90.769 -> -90.859 | Hurt |
Closest papers by meaning
- On the Limitations of Multimodal VAEs (ICLR 2022 · similarity 0.92)
- Variational Mixture-of-Experts Autoencoders for Multi-Modal Deep Generative Models (NeurIPS 2019 · similarity 0.91)
- Unity by Diversity: Improved Representation Learning for Multimodal VAEs (NeurIPS 2024 · similarity 0.90)
- Aggregation of Dependent Expert Distributions in Multimodal Variational Autoencoders (ICML 2025 · similarity 0.89)
- Integrating Multimodal Data for Joint Generative Modeling of Complex Dynamics (ICML 2024 · similarity 0.86)
- Modality-Aware SAM: Sharpness-Aware-Minimization Driven Gradient Modulation for Harmonized Multimodal Learning (NeurIPS 2025 · similarity 0.86)
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.