Research map / Architectures and efficiency / Attention and transformers
LimiX-2M: Mitigating Low-Rank Collapse and Attention Bottlenecks in Tabular Foundation Models
Yuanrui Wang, Xingxuan Zhang, Han Yu, Mingchao Hao, Gang Ren, Hao Yuan, Li Mao, Yunjia Zhang, Chun Yuan, Peng Cui
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 linear embedding with RaBEL (ablation) | 2M SNF baseline · first three layers | numerical rank: 58.41 -> 78.62 | Improved |
| replace linear embedding with RaBEL (ablation) | 2M SNF baseline · first three layers | Rank@99%: 13.94 -> 25.35 | Improved |
| replace linear embedding with RaBEL (ablation) | 2M SNF baseline · first three layers | Rank@95%: 6.73 -> 12.31 | Improved |
| replace MLP embedding with RaBEL (ablation) | Transformer · BCCO-CLS | AUC: 83.52 -> 85.04 | Improved |
| replace Periodic embedding with RaBEL (ablation) | Transformer · BCCO-CLS | AUC: 83.88 -> 85.04 | Improved |
| replace MLP embedding with RaBEL (ablation) | Transformer · BCCO-REG | RMSE: 0.4043 -> 0.3964 | Improved |
| replace Periodic embedding with RaBEL (ablation) | Transformer · BCCO-REG | R2: 0.6859 -> 0.7792 | Improved |
| replace MLP embedding with RaBEL (ablation) | MLP · GC dataset | Metric1: 0.7537 -> 0.9831 | Improved |
| replace MLP embedding with RaBEL (ablation) | MLP · MA dataset | Metric1: 0.8896 -> 0.8789 | Hurt |
| use orthogonal exponent-embedding init instead of Xavier or Kaiming (ablation) | RaBEL · BCCO-CLS | AUC (%): 89.03 | Improved |
| increase embedding dimension from 16 to 32 (ablation) | RaBEL · BCCO-CLS | AUC: 84.51 -> 85.17 | Improved |
| increase kernels from 16 to 64 (ablation) | RaBEL · BCCO-CLS | AUC: 84.20 -> 85.19 | Improved |
Closest papers by meaning
- A Closer Look at TabPFN v2: Understanding Its Strengths and Extending Its Capabilities (NeurIPS 2025 · similarity 0.89)
- Benchmarking Attention for Tabular Foundation Models (NeurIPS 2026 · similarity 0.88)
- TabDPT: Scaling Tabular Foundation Models on Real Data (NeurIPS 2025 · similarity 0.88)
- Tabular Foundation Model for Generative Modelling (NeurIPS 2026 · similarity 0.88)
- Understanding the Surprising Generalization Properties of Tabular Foundation Models (NeurIPS 2026 · similarity 0.88)
- TabArena: A Living Benchmark for Machine Learning on Tabular Data (NeurIPS 2025 · 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.
