Research map / Vision-language and multimodal / Cross-modal learning
ACMM: Aligned Cross-Modal Memory for Few-Shot Image and Sentence Matching
Yan Huang, Liang Wang
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 shared memory after cross-modal alignment (ablation) | ACMM align · Flickr30k (5000 test) | mR: 76.5 -> 80.9 | Improved |
| add shared memory after cross-modal alignment (ablation) | ACMM align · MSCOCO (5000 test) | mR: 62.0 -> 70.6 | Improved |
| add relation modeling (visual GCN and bidirectional GRU) to align (ablation) | ACMM align · Flickr30k (5000 test) | mR: 68.7 -> 76.5 | Improved |
| add relation modeling (visual GCN and bidirectional GRU) to align (ablation) | ACMM align · MSCOCO (5000 test) | mR: 49.9 -> 62.0 | Improved |
| add cross-modal alignment before memory (ablation) | ACMM memory · Flickr30k (5000 test) | mR: 5.7 -> 80.9 | Improved |
| add cross-modal alignment before memory (ablation) | ACMM memory · MSCOCO (5000 test) | mR: 1.3 -> 70.6 | Improved |
| use shared memory instead of modality-specific memory (ablation) | ACMM align + mem · Flickr30k (5000 test) | mR: 73.9 -> 80.9 | Improved |
| use shared memory instead of modality-specific memory (ablation) | ACMM align + mem · MSCOCO (5000 test) | mR: 61.2 -> 70.6 | Improved |
| use shared memory without alignment instead of modality-specific (ablation) | mem only · Flickr30k (5000 test) | mR: 0.6 -> 5.7 | No clear effect |
| ensemble with lambda 0.5 and 0.8 | ACMM · Flickr30k (1000 test) | mR: 80.9 -> 83.5 | Improved |
| ensemble with lambda 0.5 and 0.8 | ACMM · MSCOCO (5000 test) | mR: 70.6 -> 73.6 | Improved |
| replace SCAN with ACMM on 1-shot test set | SCAN · Flickr30k (5000 test), 1-shot | mR: 67.1 -> 74.2 | Improved |
Closest papers by meaning
- Cross the Gap: Exposing the Intra-modal Misalignment in CLIP via Modality Inversion (ICLR 2025 · similarity 0.87)
- SoftCLIP: Softer Cross-Modal Alignment Makes CLIP Stronger (AAAI 2024 · similarity 0.87)
- With Limited Data for Multimodal Alignment, Let the STRUCTURE Guide You (NeurIPS 2025 · similarity 0.86)
- Quantifying Cross-Modality Memorization in Vision-Language Models (NeurIPS 2025 · similarity 0.86)
- Strong Helps Weak: Directional Cross-Modal Alignment Transfer in Multi-modal LLMs (NeurIPS 2026 · similarity 0.86)
- SmartCLIP: Modular Vision-language Alignment with Identification Guarantees (CVPR 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.
