Sahaj.blogs
27 Jan 2026 · Paper summary

Paper summary: Can LLMs Clean Up Your Mess? A Survey of Application-Ready Data Preparation with LLMs

A survey of using LLMs to prepare application-ready data: cleaning, integration and enrichment, moving from rule-based pipelines to prompt-driven and agentic workflows.

Figure from the paper "Can LLMs Clean Up Your Mess? A Survey of Application-Ready Data Preparation with LLMs"
Figure from Zhou et al. (2026), licensed CC BY 4.0. Fig. 1: Application-Ready Data Preparation – Three core tasks (i.e., Data Cleaning, Integration, and Enrichment) address key sources of data inefficiency: quality issues, integration barriers, and semantic gaps.

Problem

Analytics, visualisation and decision-making need application-ready data. Traditional preparation relies on hand-crafted rules and domain-specific models, which take a lot of manual effort and expertise.

Method

The survey reviews hundreds of works and organises LLM-based preparation into three tasks (Figure 1):

  • data cleaning: standardisation, error processing, imputation
  • data integration: entity matching, schema matching
  • data enrichment: annotation, profiling

It describes the shift from rule-based, model-specific pipelines to prompt-driven, context-aware and agentic workflows.

Key results

For each task, the survey lists representative techniques with their strengths (e.g. better generalisation, semantic understanding) and their limitations. Its first limitations section explains why traditional methods are costly: heavy manual effort and dependence on expertise.

Limitations

As a survey, it maps a fast-moving area as of early 2026. A companion reading list is maintained on GitHub.

#llm#data#survey

Sources

  1. Can LLMs Clean Up Your Mess? A Survey of Application-Ready Data Preparation with LLMs (arXiv:2601.17058) — arXiv
  2. Hugging Face Daily Papers, 2026-01-27 — Hugging Face
  3. Can LLMs Clean Up Your Mess? A Survey of Application-Ready Data Preparation with LLMs (arXiv HTML full text) — arXiv
  4. Code repository (linked in the paper) — GitHub

This is a summary written from the linked public sources. Figures and numbers are as reported there.