From Rules to Predictions: Federated Tabular Learning with LLM Reasoning

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Health & Medical Research · Depth: Expert, quick

Summary

A new federated framework, "From Rules to Predictions: Federated Tabular Learning with LLM Reasoning," developed by Afsaneh Mahanipour and Hana Khamfroush for BioNLP 2026, addresses critical challenges in tabular data modeling for sensitive domains like healthcare and finance. It tackles data privacy, distributed data, and performance simultaneously, issues not fully resolved by existing methods. The framework avoids the pitfalls of direct tabular-to-text conversion for LLMs, which can expose sensitive information and be computationally expensive. Instead, it uses decision tree rules as privacy-preserving intermediaries. Each client trains a local Random Forest and shares only these rules. A global pool of rules allows an LLM to generate improved partitioning rules without accessing raw data. Clients then apply local gradient-based corrections, which are aggregated. Experiments across 12 datasets, including seven medical tasks, demonstrate that this method consistently outperforms federated baselines and achieves performance comparable to centralized models.

Key takeaway

For Machine Learning Engineers building models with sensitive, distributed tabular data in healthcare or finance, this federated LLM reasoning framework offers a robust solution. You should consider adopting this rule-based approach to enhance model performance while strictly maintaining data privacy. This method allows you to achieve near-centralized model accuracy without direct access to raw client data, mitigating significant privacy and compliance risks.

Key insights

Federated learning can leverage LLM reasoning on privacy-preserving decision rules to improve tabular data model performance without raw data exposure.

Principles

• Decision tree rules can serve as privacy-preserving intermediaries.
• LLMs can generate global partitioning rules from aggregated local rules.
• Gradient-based corrections enhance local model accuracy in federated settings.

Method

Clients train local Random Forests, share rules. An LLM combines rules to generate a global partitioning rule. Clients apply local gradient-based corrections, which are then aggregated.

In practice

• Implement Random Forests locally for initial rule extraction.
• Aggregate decision rules, not raw data, for LLM processing.
• Apply gradient-based corrections after global rule updates.

Topics

• Federated Learning
• Tabular Data
• Large Language Models
• Data Privacy
• Decision Trees
• Healthcare AI

Best for: AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• LLMs on Tabular Data with Limited Semantics: Evidence from Industrial Car Retrofit Prediction — LLMs serve best as complementary components to strong tabular machine learning baselines, not replacements, for industrial data with limited semantics.
• A federated graph learning method to realize multi-party collaboration for molecular discovery — FedLG enables privacy-preserving multi-party molecular discovery through federated graph learning and the Lanczos algorithm.
• Privacy-Aware Infrastructure in the AI-Native Era: An Asset Classification Case Study — Meta's hybrid classification system uses LLMs for ambiguity and distills stable patterns into deterministic rules for scalable, auditable privacy enforcement.
• ReSS: Learning Reasoning Models for Tabular Data Prediction via Symbolic Scaffold — ReSS combines symbolic decision trees with LLMs to generate accurate, faithful, and explainable tabular data predictions.
• Beyond the lakehouse: Fundamental's NEXUS bypasses manual ETL with a native foundation model for tabular data — NEXUS is a Large Tabular Model designed to natively understand and predict from complex, non-sequential business data.
• Responsible Federated LLMs via Safety Filtering and Constitutional AI — Integrating safety filters and Constitutional AI significantly improves FedLLM safety against harmful content.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.