CLEAR: Cost-Efficient and Reliability-Aware Tabular Data Cleaning with Calibrated Large Language Models — Reproducibility Package

Complete reproducibility package for the paper "CLEAR: Cost-Efficient and Reliability-Aware Tabular Data Cleaning with Calibrated Large Language Models". Contains the detector, self-consistency LLM repair client, calibration and selective-repair implementation, dataset preparation scripts (HoloClean Hospital/Flight, UCI Adult), benchmark subsets, all configuration files, and the raw per-run results (dirty tables, proposals, results.json) for 17 experimental runs. All experiments ran on a laptop CPU with locally served open-weight models at zero monetary cost.