Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2602.06205 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914308787535872 |
|---|---|
| author | Achara, Akshit Gaintseva, Tatiana Mahaut, Mateo Chakraborty, Pritish Johansson, Viktor Stenby Barsbey, Melih Rodolà, Emanuele Crisostomi, Donato |
| author_facet | Achara, Akshit Gaintseva, Tatiana Mahaut, Mateo Chakraborty, Pritish Johansson, Viktor Stenby Barsbey, Melih Rodolà, Emanuele Crisostomi, Donato |
| contents | The Platonic Representation Hypothesis suggests that independently trained neural networks converge to increasingly similar latent spaces. However, current strategies for mapping these representations are inherently pairwise, scaling quadratically with the number of models and failing to yield a consistent global reference. In this paper, we study the alignment of $M \ge 3$ models. We first adapt Generalized Procrustes Analysis (GPA) to construct a shared orthogonal universe that preserves the internal geometry essential for tasks like model stitching. We then show that strict isometric alignment is suboptimal for retrieval, where agreement-maximizing methods like Canonical Correlation Analysis (CCA) typically prevail. To bridge this gap, we finally propose Geometry-Corrected Procrustes Alignment (GCPA), which establishes a robust GPA-based universe followed by a post-hoc correction for directional mismatch. Extensive experiments demonstrate that GCPA consistently improves any-to-any retrieval while retaining a practical shared reference space. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_06205 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Multi-Way Representation Alignment Achara, Akshit Gaintseva, Tatiana Mahaut, Mateo Chakraborty, Pritish Johansson, Viktor Stenby Barsbey, Melih Rodolà, Emanuele Crisostomi, Donato Machine Learning Artificial Intelligence The Platonic Representation Hypothesis suggests that independently trained neural networks converge to increasingly similar latent spaces. However, current strategies for mapping these representations are inherently pairwise, scaling quadratically with the number of models and failing to yield a consistent global reference. In this paper, we study the alignment of $M \ge 3$ models. We first adapt Generalized Procrustes Analysis (GPA) to construct a shared orthogonal universe that preserves the internal geometry essential for tasks like model stitching. We then show that strict isometric alignment is suboptimal for retrieval, where agreement-maximizing methods like Canonical Correlation Analysis (CCA) typically prevail. To bridge this gap, we finally propose Geometry-Corrected Procrustes Alignment (GCPA), which establishes a robust GPA-based universe followed by a post-hoc correction for directional mismatch. Extensive experiments demonstrate that GCPA consistently improves any-to-any retrieval while retaining a practical shared reference space. |
| title | Multi-Way Representation Alignment |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2602.06205 |