Research map / Architectures and efficiency / Attention and transformers
Breaking the Correlation Plateau: On the Optimization and Capacity Limits of Attention-Based Regressors
Jingquan Yan, Yuwei Miao, Peiran Yu, Junzhou Huang
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 |
|---|---|---|---|
| add ECA to ALMT (with dispersion-normalized PCC loss) | ALMT · MOSI | PCC: 0.783 -> 0.806 | Improved |
| add a plain PCC loss to ALMT | ALMT · MOSI | F1: 0.851 -> 0.834 | Hurt |
| add ECA to EGN | EGN · 10xProteomic (3-fold) | PCC@M: 0.629 -> 0.716 | Improved |
| add ECA to FT-Transformer | FT-Transformer · Appliance (UCI) | PCC: 0.556 -> 0.598 | Improved |
| add ECA to FT-Transformer | FT-Transformer · Online News (UCI) | PCC: 0.408 -> 0.420 | Improved |
| add ECA to FT-Transformer | FT-Transformer · Superconductivity (UCI) | PCC: 0.920 -> 0.930 | Improved |
| add DATS (dispersion-aware temperature softmax) (ablation) | ECA · Appliance (UCI) | PCC: 0.561 -> 0.598 | Improved |
| add SRA (scaled residual aggregation) (ablation) | ECA · Online News (UCI) | PCC: 0.410 -> 0.420 | Improved |
| add DNPL (dispersion-normalized PCC loss) (ablation) | ECA · Superconductivity (UCI) | PCC: 0.922 -> 0.930 | Improved |
| add SRA to ECA (ablation) | ECA · Appliance (UCI) | MSE (x10^3): 5.994 -> 5.790 | Improved |
| add DATS to ECA (ablation) | ECA · Online News (UCI) | PCC: 0.418 -> 0.420 | No clear effect |
Closest papers by meaning
- Attention-based PCA (NeurIPS 2026 · similarity 0.87)
- In-Context Linear Regression Demystified: Training Dynamics and Mechanistic Interpretability of Multi-Head Softmax Attention (ICML 2025 · similarity 0.87)
- Attention layers provably solve single-location regression (ICLR 2025 · similarity 0.85)
- Degrees of Freedom for Linear Attention: Distilling Softmax Attention with Optimal Feature Efficiency (NeurIPS 2025 · similarity 0.85)
- Sparse Attention as Compact Kernel Regression (NeurIPS 2026 · similarity 0.85)
- Training Dynamics of In-Context Learning in Linear Attention (ICML 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.
