Ensembling LLM-Induced Decision Trees for Explainable and Robust Error Detection

2025/12/08 by Mengqi Wang, Jianwei Wang, Wang, Mengqi +13 · 1 citation
Biochemistry, Genetics and Molecular Biology · Computer Science · #cs.CL

paper · pdf · doi:10.48550/arxiv.2512.07246

15 pages, 7 figures. Accepted at KDD 2026

arxiv created 2026/07/29 · arxiv updated 2026/07/30

Abstract

Error detection (ED), which aims to identify incorrect or inconsistent cell values in tabular data, is important for ensuring data quality. Recent state-of-the-art ED methods leverage the pre-trained knowledge and semantic capability of large language models (LLMs) to directly label whether a cell is erroneous. However, this LLM-as-a-labeler pipeline produces predictions through an implicit black-box process with limited traceability and explicit justification, and relies on stochastic single-pass inference, resulting in inconsistent and insufficiently robust detections across contexts. To address these limitations, we propose an LLM-as-an-inducer framework that uses an LLM to induce a decision tree for ED, termed TreeED, and ensembles multiple such trees for consensus detection, termed ForestED. Based on prompts derived from data context, decision tree specifications, and output requirements, TreeED queries the LLM to induce a decision tree skeleton whose root-to-leaf paths specify the stepwise procedure for evaluating a sample. Each tree contains three types of nodes: (1) rule nodes that perform simple validation checks, such as format or range constraints; (2) Graph Neural Network (GNN) nodes that capture complex patterns, such as functional dependencies; and (3) leaf nodes that output the final decision as error or clean. ForestED employs uncertainty sampling to obtain multiple informative row subsets and constructs a decision tree for each subset using TreeED. It then applies an Expectation-Maximization-based algorithm to jointly estimate tree reliability and optimize the consensus ED prediction. Experiments demonstrate that our methods are accurate, explainable, and robust, achieving an average F1-score improvement of 16.1% over the best baseline.

Citations

• Walk the Talk? Measuring the Faithfulness of Large Language Model Explanations
• Exploring LLM Agents for Cleaning Tabular Machine Learning Datasets
• Few-shot LLM Synthetic Data with Distribution Matching
• CORTEX: A Cost-Sensitive Rule and Tree Extraction Method
• On LLM-Enhanced Mixed-Type Data Imputation with High-Order Message Passing
• GPTree: Towards Explainable Decision-Making via LLM-powered Decision Trees
• Schema Matching with Large Language Models: an Experimental Study
• The ALCHEmist: Automated Labeling 500x CHEaper Than LLM Data Annotators
• On LLMs-Driven Synthetic Data Generation, Curation, and Evaluation: A Survey
• Match, Compare, or Select? An Investigation of Large Language Models for Entity Matching
• Jellyfish: A Large Language Model for Data Preprocessing
• Distributing Accountability, Not Capability: Phase Separation and the LLM Workflow Quadrant in Autonomous AI Agent Architectures
• Can Foundation Models Wrangle Your Data?
• Handling Missing Data with Graph Representation Learning
• Inductive Representation Learning on Large Graphs

Cited by

• Interpretable Column Annotation with LLM-Symbolized Decision Process Materialization
• Empowering Tabular Data Preparation with Language Models: Why and How?

Related

• LLM Meeting Decision Trees on Tabular Data
• Talking Trees: Reasoning-Assisted Induction of Decision Trees for Tabular Data
• Improving the Validity of Decision Trees as Explanations
• "Oh LLM, I'm Asking Thee, Please Give Me a Decision Tree": Zero-Shot Decision Tree Induction and Embedding with Large Language Models
• Optimized Feature Generation for Tabular Data via LLMs with Decision Tree Reasoning