Identifiable alignment of unpaired representations
stat.ML, cs.AI, cs.LG, stat.ME
Submitted: 2026-11-08
Updated: 2026-11-08
Code: https://github.com/mustass/identifiable-alignment
Terminology
Sources
- Latent Space Oddity: on the Curvature of Deep Generative Models
- LeJEPA: Provable and Scalable Self-Supervised Learning Without the Heuristics
- Proving and Improving the Platonic Representation Hypothesis with JEPAs
- Word Translation Without Parallel Data
- Similarity and Matching of Neural Network Representations
- Sinkhorn Distances: Lightspeed Computation of Optimal Transportation Distances
- mini-vec2vec: Scaling Universal Geometry Alignment with Linear Transformations
- Only Bayes should learn a manifold (on the estimation of differential geometric structure from data)
- Gaussian Error Linear Units (GELUs)
- Auto-Encoding Variational Bayes
- Note on computational complexity of the Gromov-Wasserstein distance
- Efficient Estimation of Word Representations in Vector Space
- Relative representations enable zero-shot latent space communication
- Fast and Robust Comparison of Probability Measures in Heterogeneous Spaces
- Shared Geometry As A Rosetta Stone: Cross-Modal Alignment Without Paired Data
- Strongly Isomorphic Neural Optimal Transport Across Incomparable Spaces
- Decoder ensembling for learned latent geometries
- Identifying Metric Structures of Deep Latent Variable Models
- Disentangled Representation Learning with the Gromov-Monge Gap
- Indeterminacy in Generative Models: Characterization and Strong Identifiability
Related papers
- Uniform-in-Time Weak Propagation-of-Chaos in Shallow Neural Networks
- Diffusion Models Observe Only Gradients: A Geometric Perspective on Score Matching Errors
- Quantifying Error Propagation and Model Collapse in Diffusion Models
- Behavior of prediction performance metrics with rare events
- Statistical Inference and Learning for Shapley Additive Explanations (SHAP)
- Control Variate Design for Score-Function Gradient Estimators in Importance-Weighted Variational Inference