Research map / Probabilistic and scientific modeling / Molecules and proteins
Domain Expansion: A Latent Space Construction Framework for Multi-Task Learning
Chi-Yao Huang, Khoa Vo, Aayush Atul Verma, Duo Lu, Yezhou Yang
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 |
|---|---|---|---|
| replace shared latent space with orthogonal pooling (Domain Expansion) | weighted-sum baseline · ShapeNet Objective Set 1 | Spearman (azimuth): 0.41 -> 0.95 | Improved |
| use Domain Expansion instead of gradient-balancing MTL (FAMO, Nash-MTL, IMTL) | FAMO/Nash-MTL/IMTL · ShapeNet Objective Set 1 | V-score (category): 0.00/0.00/0.39 -> 0.99 | Improved |
| use Domain Expansion on SupCon classification objective | baselines · ShapeNet Objective Set 2 | V-score (id): 0.00 -> 0.96 | Improved |
| use Domain Expansion on classification accuracy under Set 2 | baseline · ShapeNet Objective Set 2 | accuracy (category): 0.99 -> 0.98 | No clear effect |
| use Domain Expansion for concept composition | baseline · ShapeNet Objective Set 1 | cosine similarity: 0.22 -> 0.95 | Improved |
| use Domain Expansion on MPIIGaze | baseline · MPIIGaze Objective Set 1 | Spearman (gaze y): 0.48 -> 0.92 | Improved |
| add Hungarian alignment of eigenvectors across epochs (ablation) | Domain Expansion · ShapeNet | eigenvector cosine similarity (elevation): 0.04 -> 0.99 | Improved |
| add Hungarian alignment for ID concept (ablation) | Domain Expansion · ShapeNet | eigenvector cosine similarity (ID): 0.98 -> 0.99 | No clear effect |
Closest papers by meaning
- Latent Functional Maps: a spectral framework for representation alignment (NeurIPS 2024 · similarity 0.86)
- Latent Zoning Network: A Unified Principle for Generative Modeling, Representation Learning, and Classification (NeurIPS 2025 · similarity 0.86)
- Modeling Multi-Task Model Merging as Adaptive Projective Gradient Descent (ICML 2025 · similarity 0.86)
- No Task Left Behind: Isotropic Model Merging with Common and Task-Specific Subspaces (ICML 2025 · similarity 0.86)
- Connecting Neural Models Latent Geometries with Relative Geodesic Representations (NeurIPS 2025 · similarity 0.85)
- Disentangling Representations through Multi-task Learning (ICLR 2025 · similarity 0.85)
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.
