# STEER: Reducing Inference Cost in Relational Foundation Models through Semantically Informed Sampling

arXiv: https://arxiv.org/abs/2610.00907 · HTML: https://mlsanity.com/paper/2610.00907/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 22 of its 38 references that are in the map.
- First posted on arXiv 1 Oct 2026.

## Builds on (most cited first, 22 of 22)

- [Inductive Representation Learning on Large Graphs](https://mlsanity.com/paper/1706.02216/) (2017)
- [XGBoost: A Scalable Tree Boosting System](https://mlsanity.com/paper/1603.02754/) (2016)
- [Fast Inference from Transformers via Speculative Decoding](https://mlsanity.com/paper/2211.17192/) (2022)
- [Graph Convolutional Neural Networks for Web-Scale Recommender Systems](https://mlsanity.com/paper/1806.01973/) (2018)
- [Large Language Models are Effective Text Rankers with Pairwise Ranking Prompting](https://mlsanity.com/paper/2306.17563/) (2023)
- [TabICL: A Tabular Foundation Model for In-Context Learning on Large Data](https://mlsanity.com/paper/2502.05564/) (2025)
- [CodeS: Towards Building Open-source Language Models for Text-to-SQL](https://mlsanity.com/paper/2402.16347/) (2024)
- [TabICLv2: A better, faster, scalable, and open tabular foundation model](https://mlsanity.com/paper/2602.11139/) (2026)
- [RelBench: A Benchmark for Deep Learning on Relational Databases](https://mlsanity.com/paper/2407.20060/) (2024)
- [Squeezed Attention: Accelerating Long Context Length LLM Inference](https://mlsanity.com/paper/2411.09688/) (2024)
- [RelGNN: Composite Message Passing for Relational Deep Learning](https://mlsanity.com/paper/2502.06784/) (2025)
- [Relational Transformer: Toward Zero-Shot Foundation Models for Relational Data](https://mlsanity.com/paper/2510.06377/) (2025)
- [Large Language Models are Good Relational Learners](https://mlsanity.com/paper/2506.05725/) (2025)
- [Schema Matching with Large Language Models: an Experimental Study](https://mlsanity.com/paper/2407.11852/) (2024)
- [RelBench v2: A Large-Scale Benchmark and Repository for Relational Data](https://mlsanity.com/paper/2602.12606/) (2026)
- [PluRel: Synthetic Data unlocks Scaling Laws for Relational Foundation Models](https://mlsanity.com/paper/2602.04029/) (2026)
- [KumoRFM-2: Scaling Foundation Models for Relational Learning](https://mlsanity.com/paper/2604.12596/) (2026)
- [Tackling prediction tasks in relational databases with LLMs](https://mlsanity.com/paper/2411.11829/) (2024)
- [No Need to Train Your RDB Foundation Model](https://mlsanity.com/paper/2602.13697/) (2026)
- [Transformers Meet Relational Databases](https://mlsanity.com/paper/2412.05218/) (2024)
- [Predictive Query Language: A Domain-Specific Language for Predictive Modeling on Relational Databases](https://mlsanity.com/paper/2602.09572/) (2026)
- [An LLM-Guided Query-Aware Inference System for GNN Models on Large Knowledge Graphs](https://mlsanity.com/paper/2603.04545/) (2026)

## Abstract (arXiv)

Relational foundation models (RFMs) are pretrained once on a collection of relational databases and prediction tasks, and then applied zero-shot to previously unseen databases and tasks. To make a prediction for a target row, an RFM samples a neighborhood of rows linked to that row through foreign keys and uses this neighborhood as its inference context. Lowering inference cost is an important goal for any foundation model, and for RFMs this cost grows with the size of the context. The simplest ways to shrink the context is to drop some of the sampled rows, but this ignores the semantics of the database schema, so it is as likely to discard informative rows as uninformative ones. We propose STEER, a sampling approach that shrinks the inference context by concentrating it on the tables most relevant to the prediction task at hand. STEER obtains relevance information by prompting a large language model to rank the foreign-key edges of the database schema into relevance tiers for the given task, and then maps each tier to a probability of following that edge during traversal. Because the ranking uses only the schema, it is computed once per task and reused across all subsequent predictions, amortizing its cost. We evaluate STEER on three state-of-the-art RFMs (RT, RT-J, and Griffin) and show that it reduces inference context size by about 40% on average while maintaining, and in some cases improving, prediction accuracy.