Research map / Architectures and efficiency / Pruning and efficient tuning
No Need to Train Your RDB Foundation Model
Linjie Xu, Yanlin Zhang, Quan Gan, Minjie Wang, David Wipf
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 cross-column dense encoder mixing in front of the ICL decoder with many uninformative columns (ablation) | TabPFNv2 ICL decoder · synthetic classification | accuracy | Hurt |
| add cross-column dense encoder when added columns are informative (ablation) | TabPFNv2 ICL decoder · synthetic classification | accuracy | No clear effect |
| swap dense parameterized encoder for JUICE-style parameter-free encoder under ICL | KumoRFM vs RDBLearn · RelBench-v1 | AUC difference | Improved |
| use JUICE encoder with supervised head instead of dense encoder | RelGT vs AutoGluon on JUICE · RelBench-v1 | AUC difference | Hurt |
| vary base tabular model (TabPFN-v2, v2.5, LimiX) and hops H in {2,3} (ablation) | RDBLearn · RelBench-v1 classification | AUC | No clear effect |
| replace random GNN encoder with JUICE (no cross-column mixing) (ablation) | RandGNN · RelBench-v1 classification | AUC | Improved |
| use RDBLearn instead of RandGNN | 4DBInfer Retailrocket CVR · Retailrocket CVR | AUC: 0.6621 -> 0.8469 | Improved |
| use RDBLearn instead of RandGNN | 4DBInfer Outbrain CTR-100K · Outbrain CTR-100K | AUC: 0.4929 -> 0.5447 | Improved |
| train RT without pre-training on each test RDB | RT · RelBench-v1 classification | AUC: 70.1 -> 66.7 | Hurt |
Closest papers by meaning
- Griffin: Towards a Graph-Centric Relational Database Foundation Model (ICML 2025 · similarity 0.87)
- RDB2G-Bench: A Comprehensive Benchmark for Automatic Graph Modeling of Relational Databases (NeurIPS 2025 · similarity 0.85)
- Joint Relational Database Generation via Graph-Conditional Diffusion Models (NeurIPS 2025 · similarity 0.85)
- TabDPT: Scaling Tabular Foundation Models on Real Data (NeurIPS 2025 · similarity 0.85)
- 4DBInfer: A 4D Benchmarking Toolbox for Graph-Centric Predictive Modeling on RDBs (NeurIPS 2024 · similarity 0.85)
- Position: Relational Deep Learning - Graph Representation Learning on Relational Databases (ICML 2024 · 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.
