Research map / Learning theory and algorithms / Prediction and decision losses
GOTabPFN: From Feature Ordering to Compact Tokenization for Tabular Foundation Models on High-Dimensional Data
Al Zadid Sultan Bin Habib, Md Younus Ahamed, Prashnna Kumar Gyawali, Gianfranco Doretto, Donald Adjeroh
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 NSC compression (GO-LR+NSC-pSP) before TabPFN-2.5 (ablation) | GOTabPFN · Colon (ablation) | accuracy (%): 86.85 -> 88.18 | Improved |
| add GO-LR ordering instead of random order (ablation) | GOTabPFN · Colon (ablation) | accuracy (%): 84.21 -> 88.18 | Improved |
| add GO-LR ordering instead of identity order (ablation) | GOTabPFN · Colon (ablation) | accuracy (%): 81.67 -> 88.18 | Improved |
| use TabPFN-2.5 head instead of logistic regression (ablation) | GOTabPFN · Colon (ablation) | accuracy (%): 82.67 -> 88.18 | Improved |
| use transition-aware segmentation instead of uniform segmentation (ablation) | GOTabPFN · Colon (ablation) | accuracy (%): 80.00 -> 88.18 | Improved |
| add GO-LR ordering instead of mean-pool NSC without ordering (ablation) | GOTabPFN · Colon (ablation) | accuracy (%): 64.59 -> 88.18 | Improved |
| replace greedy GO-LR with simulated annealing ordering (ablation) | GOTabPFN · Colon (ordering method) | accuracy: 0.8818 -> 0.8405 | Hurt |
| replace GO-LR with ant colony ordering (ablation) | GOTabPFN · Colon (ordering method) | accuracy: 0.8818 -> 0.8595 | Hurt |
| replace GO-LR with Christofides ordering (ablation) | GOTabPFN · Colon (ordering method) | accuracy: 0.8818 -> 0.8364 | Hurt |
| set GO-LR cluster size k to 10 instead of 3 (ablation) | GOTabPFN · Colon (cluster-size sweep) | accuracy (%): 82.72 -> 88.18 | Improved |
| increase GO-LR cluster size k from 10 to 15 (ablation) | GOTabPFN · Colon (cluster-size sweep) | accuracy (%): 88.18 -> 82.69 | Hurt |
| add GO-LR + NSC in front of best original foundation-model head | TabPFN-2.5 · GLI-85 (HDLSS) | accuracy (%): 89.66 -> 93.82 | Improved |
Closest papers by meaning
- A Closer Look at TabPFN v2: Understanding Its Strengths and Extending Its Capabilities (NeurIPS 2025 · similarity 0.89)
- TabPFN Unleashed: A Scalable and Effective Solution to Tabular Classification Problems (ICML 2025 · similarity 0.88)
- Benchmarking Attention for Tabular Foundation Models (NeurIPS 2026 · similarity 0.87)
- TabDPT: Scaling Tabular Foundation Models on Real Data (NeurIPS 2025 · similarity 0.87)
- TabM: Advancing tabular deep learning with parameter-efficient ensembling (ICLR 2025 · similarity 0.87)
- TabPrep: Closing the Feature Engineering Gap in Tabular Benchmarks (NeurIPS 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.
