Research map / Vision-language and multimodal / Cross-modal learning
Learning a Recurrent Residual Fusion Network for Multimodal Matching
Yu Liu, Yanming Guo, Erwin M. Bakker, Michael S. Lew
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 RRF block to the plain matching network | baseline network · Flickr30K test image-to-text | R@1: 45.0 -> 47.6 | Improved |
| add RRF block to the plain matching network | baseline network · Flickr30K test text-to-image | R@1: 33.6 -> 35.4 | Improved |
| increase recurrent steps from T=3 to T=4 (ablation) | RRF-Net · Flickr30K test image-to-text | R@1: 47.6 -> 46.2 | Hurt |
| increase recurrent steps from T=1 to T=3 (ablation) | RRF-Net · Flickr30K test image-to-text | R@1: 46.4 -> 47.6 | Improved |
| add fusion module to recurrent residual model (ablation) | RRF-Net · Flickr30K test image-to-text | R@1: 45.8 -> 47.6 | Improved |
| use convolutional fusion instead of sum fusion (ablation) | RRF-Net · Flickr30K test image-to-text | R@1: 47.1 -> 47.6 | Improved |
| use bi-rank loss instead of bi-directional loss (ablation) | baseline network · Flickr30K test image-to-text | R@1: 43.4 -> 45.0 | Improved |
| use bi-rank loss instead of bi-directional loss (ablation) | RRF-Net · Flickr30K test image-to-text | R@1: 46.4 -> 47.6 | Improved |
| ensemble four RRF-Net models with T=1..4 | RRF-Net · Flickr30K test image-to-text | R@1: 47.6 -> 50.8 | Improved |
| ensemble four RRF-Net models with T=1..4 | RRF-Net · Flickr30K test text-to-image | R@1: 35.4 -> 37.6 | Improved |
| add RRF blocks, cross-dataset transfer | baseline network · Train Flickr30K, test MSCOCO image-to-text | R@1: 23.7 -> 24.8 | Improved |
| use ResNet-152 instead of VGG-19 visual features | RRF-Net · Flickr30K test image-to-text | R@1: 42.1 -> 47.6 | Improved |
Closest papers by meaning
- Noisy Correspondence Rectification via Asymmetric Similarity Learning (AAAI 2025 · similarity 0.87)
- Recurrence-Enhanced Vision-and-Language Transformers for Robust Multimodal Document Retrieval (CVPR 2025 · similarity 0.87)
- Multimodal Distribution Matching for Vision-Language Dataset Distillation (CVPR 2026 · similarity 0.87)
- Same Concept, Different Directions: Cross-Modal Feature Heterogeneity in Sparse Autoencoders (NeurIPS 2026 · similarity 0.86)
- Reconstruction Alignment Improves Unified Multimodal Models (ICLR 2026 · similarity 0.86)
- LoMo: Local Modality Substitution for Deeper Vision-Language Fusion (NeurIPS 2026 · 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.
