Research map / Vision-language and multimodal / CLIP and zero-shot
Latent Space Translation via Semantic Alignment
Valentino Maiorca, Luca Moschella, Antonio Norelli, Marco Fumero, Francesco Locatello, Emanuele Rodolà
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 |
|---|---|---|---|
| translate with ortho instead of using absolute encodings | latent translation · CIFAR-10 stitching | accuracy: 0.16 ± 0.22 -> 0.93 ± 0.04 | Improved |
| translate with ortho instead of relative representations | latent translation · CIFAR-100 fine stitching | accuracy: 0.30 ± 0.24 -> 0.71 ± 0.09 | Improved |
| translate with ortho instead of relative representations | latent translation · TREC | accuracy: 0.36 ± 0.13 -> 0.79 ± 0.11 | Improved |
| translate with ortho instead of relative representations | latent translation · IMDB | accuracy: 0.51 ± 0.02 -> 0.59 ± 0.04 | Improved |
| replace unconstrained linear with l-ortho (orthogonalized linear) (ablation) | latent translation · MNIST stitching | accuracy: 0.89 ± 0.08 -> 0.81 ± 0.11 | Hurt |
| replace unconstrained linear with l-ortho (ablation) | latent translation · CIFAR-10 stitching | accuracy: 0.88 ± 0.11 -> 0.90 ± 0.09 | Improved |
| replace linear with l-ortho (ablation) | latent translation · TREC | accuracy: 0.74 ± 0.25 -> 0.57 ± 0.25 | Hurt |
| replace ortho with affine (broader class) (ablation) | latent translation · TREC | accuracy: 0.79 ± 0.11 -> 0.82 ± 0.12 | Improved |
| replace ortho with affine (ablation) | latent translation · CIFAR-10 | accuracy: 0.93 ± 0.04 -> 0.92 ± 0.05 | No clear effect |
| use L2 normalization instead of standard scaling (ablation) | linear translation · MNIST stitching | accuracy: 0.89 ± 0.08 -> 0.61 ± 0.23 | Hurt |
| use L2 normalization instead of standard scaling (ablation) | linear translation · TREC | accuracy: 0.74 ± 0.25 -> 0.44 ± 0.20 | Hurt |
| use L2 normalization instead of standard scaling (ablation) | ortho translation · TREC | accuracy: 0.79 ± 0.11 -> 0.77 ± 0.12 | No clear effect |
Closest papers by meaning
- Escaping Plato's Cave: Towards the Alignment of 3D and Text Latent Spaces (CVPR 2025 · similarity 0.88)
- Connecting Neural Models Latent Geometries with Relative Geodesic Representations (NeurIPS 2025 · similarity 0.88)
- From Bricks to Bridges: Product of Invariances to Enhance Latent Space Communication (ICLR 2024 · similarity 0.88)
- Latent Functional Maps: a spectral framework for representation alignment (NeurIPS 2024 · similarity 0.88)
- Joint‑Embedding vs Reconstruction: Provable Benefits of Latent Space Prediction for Self‑Supervised Learning (NeurIPS 2025 · similarity 0.87)
- LatentLens: Revealing Highly Interpretable Visual Tokens in LLMs (ICML 2026 · similarity 0.87)
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.
