Research map / Image generation / Autoregressive and latent generation
Uncorrelation and Evenness: a New Diversity-Promoting Regularizer
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| add uniform eigenvalue regularizer (UER) | DML · Stanford-Cars | Precision@10 (%): 53.1 -> 58.2 | Helped |
| add UER | DML · MIMIC-III | Precision@10 (%): 72.5 -> 75.4 | Helped |
| add UER | DML · Birds | Precision@10 (%): 55.9 -> 59.4 | Helped |
| use UER instead of mutual-angle regularizer | DML · Birds | Precision@10 (%): 58.2 -> 59.4 | Helped |
| replace L2, L1, low-rank, dropout regularizers with diversity-promoting ones | DML · MIMIC-III | Precision@10 (%): 72.9 -> 74.3 | Mixed |
| add UER | DML · MIMIC-III infrequent diseases | Precision@10 (%): 64.2 -> 70.7 | Helped |
| add UER | DML · MIMIC-III frequent diseases | Precision@10 (%): 77.6 -> 78.3 | No clear effect |
| add UER to reduce number of components | DML · Stanford-Cars | components needed: 300 -> 100 | Helped |
| add UER | DML · MIMIC-III, 200 components | average pairwise correlation: 0.73 -> 0.57 | Helped |
| add UER | DML · MIMIC-III, 200 components | KL divergence to uniform: 3.54 -> 2.92 | Helped |
| add UER | DML · MIMIC-III training time | relative runtime increase: +11.2% | No clear effect |
| add UER | BIDAF · CNN test | accuracy (%): 76.94 -> 77.27 | Helped |
Closest papers by meaning
- Metric Space Magnitude for Evaluating the Diversity of Latent Representations (NeurIPS 2024 · similarity 0.85)
- Stabilize the Latent Space for Image Autoregressive Modeling: A Unified Perspective (NeurIPS 2024 · similarity 0.85)
- Linear combinations of latents in generative models: subspaces and beyond (ICLR 2025 · similarity 0.84)
- Compress to Impress: Efficient LLM Adaptation Using a Single Gradient Step on 100 Samples (NeurIPS 2025 · similarity 0.84)
- CORAL: Disentangling Latent Representations in Long-Tailed Diffusion (NeurIPS 2025 · similarity 0.84)
- Mapping the Multiverse of Latent Representations (ICML 2024 · 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.