aiqbench analysis
The paper proposes RelAgent, where an LLM agent designs SQL feature programs and selects a classical model during a search phase, then runs the resulting pipeline without further LLM calls at inference. That design trades the flexibility of end-to-end neural relational models for deterministic, interpretable predictors that run on standard database systems. The key open question for readers is whether these gains hold on benchmarks beyond those the authors chose, and at what cost in search-phase compute.
RelAgent: LLM Agents as Data Scientists for Relational Learning
arXiv:2605.07840v2 Announce Type: replace Abstract: Relational learning is a challenging problem that has motivated a wide range of approaches, including graph-based models (e.g., graph neural networks, graph transformers), tabular methods (e.g., tabular foundation models), and sequence-based approaches (e.g., large language models), each with its own advantages and limitations. We propose RelAgent, an LLM-based autonomous data scientist for relational learning, which operates in two phases. In the search phase, an LLM agent uses database, validation, and evaluation workspace tools to construct SQL feature programs and select a predictive model. In the inference phase, the resulting program is executed without further LLM calls. The final predictor consists of SQL queries and a classical model, enabling fast, deterministic, and intrinsically interpretable predictions: features are human-readable queries, and predictions depend only on the resulting query-defined feature map, enabling scalable deployment using standard database systems.
Abstract:Relational learning is a challenging problem that has motivated a wide range of approaches, including graph-based models (e.g., graph neural networks, graph transformers), tabular methods (e.g., tabular foundation models), and sequence-based approaches (e.g., large language models), each with its own advantages and limitations. We propose RelAgent, an LLM-based autonomous data scientist for relational learning, which operates in two phases. In the search phase, an LLM agent uses database, validation, and evaluation workspace tools to construct SQL feature programs and select a predictive model. In the inference phase, the resulting program is executed without further LLM calls. The final predictor consists of SQL queries and a classical model, enabling fast, deterministic, and intrinsically interpretable predictions: features are human-readable queries, and predictions depend only on the resulting query-defined feature map, enabling scalable deployment using standard database systems.
Machine Learning (cs.LG)
arXiv:2605.07840 [cs.LG]
(or arXiv:2605.07840v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2605.07840 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 227 words
Related coverage
Targeted Review for AI-Assisted Biodiversity Surveys: Active Continuous-Score Occupancy Modeling
arXiv · Sep 23, 2026
Artificial intelligences and human scientists exhibit complementary strengths in theory building
arXiv · Sep 29, 2026
The Continual Constitutional Convention: How language models revise the documents that govern them
AI Alignment Forum · Oct 13, 2026
Personalized Auto-Research: Towards a True AI Co-Scientist
arXiv · Aug 18, 2026