Research map / Label-efficient and robust learning / Recognition and long-tail
Multi-Way Representation Alignment
Akshit Achara, Tatiana Gaintseva, Matéo Mahaut, Pritish Chakraborty, Viktor Stenby Johansson, Melih Barsbey, Emanuele Rodolà, Donato Crisostomi
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 |
|---|---|---|---|
| apply geometry-corrected alignment (GCPA) on top of GPA | GPA · MASSIVE clustering | ARI: 0.198 -> 0.300 | Improved |
| apply GCPA instead of GCCA | GCCA · MASSIVE clustering | NMI: 0.508 -> 0.617 | Improved |
| align with GPA instead of no alignment | NA · MASSIVE clustering | ARI: 0.198 -> 0.198 | No clear effect |
| apply GCPA instead of GCCA, 3 languages | GCCA · TED-Multi (M=3) | rank-1 accuracy, worst pair: 0.509 -> 0.532 | Improved |
| apply GCPA instead of GCCA, 10 languages | GCCA · TED-Multi (M=10) | rank-1 accuracy, worst pair: 0.288 -> 0.314 | Improved |
| align with GPA universe instead of pairwise maps, 10 languages | PW · TED-Multi (M=10) | rank-1 accuracy, average: 0.430 -> 0.433 | Improved |
| add GCPA correction to GPA for multimodal agreement | GPA · Flickr8k triplets | matched cross-modal distance: 0.729 -> 0.560 | Improved |
| add GCPA correction to GPA | GPA · Flickr8k triplets | Gamma90: 0.760 -> 0.853 | Improved |
| apply GCPA instead of GCCA under noisy correspondences | GCCA · Flickr30k noisy | R@1: 0.607 -> 0.613 | Improved |
| add GCPA correction when fragile pair has 50% Canny corruption (ablation) | GPA · CIFAR-100 fragile pair | probing accuracy: 0.719 -> 0.643 | Hurt |
| add GCPA correction when fragile pair has 75% Canny corruption (ablation) | GPA · CIFAR-100 fragile pair | probing accuracy: 0.670 -> 0.568 | Hurt |
| add new model with GPA-ADD instead of refitting universe (ablation) | GPA-REFIT · CIFAR-100 continual addition | mean difference from refitting: -0.002 | No clear effect |
Closest papers by meaning
- Latent Functional Maps: a spectral framework for representation alignment (NeurIPS 2024 · similarity 0.87)
- Towards a learning theory of representation alignment (ICLR 2025 · similarity 0.86)
- Escaping Plato's Cave: Towards the Alignment of 3D and Text Latent Spaces (CVPR 2025 · similarity 0.86)
- Connecting Neural Models Latent Geometries with Relative Geodesic Representations (NeurIPS 2025 · similarity 0.86)
- Aligning Embeddings and Geometric Random Graphs: Informational Results and Computational Approaches for the Procrustes-Wasserstein Problem (NeurIPS 2024 · similarity 0.85)
- Functional Alignment Can Mislead: Examining Model Stitching (ICML 2025 · similarity 0.84)
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.
