Multi-Objective Agentic Rewrites for Unstructured Data Processing
Lindsey Linxi Wei 1† , Shreya Shankar 2† , Sepanta Zeighami 2 , Yeounoh Chung 3 , Fatma Ozcan 3 , Aditya G. Parameswaran 2 · 2025
Read Multi-Objective Agentic Rewrites for Unstructured Data Processing by Lindsey Linxi Wei 1† , Shreya Shankar 2† , Sepanta Zeighami 2 , Yeounoh Chung 3 , Fatma Ozcan 3 , Aditya G. Parameswaran 2 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-02289.txt.
Introduction
DocETL is an open-source system that runs language-model pipelines over messy, unstructured document collections, used in fields from journalism to law and medicine. Lindsey Wei and five co-authors extend it in 2025 with agentic rewrites that optimize these pipelines against several goals at once.
A year after release, the system had gathered thousands of users and GitHub stars across many domains. The new work lets an agent rewrite a data-processing pipeline and judge the rewrite on multiple objectives, like cost and accuracy together. Automated optimization of document pipelines could make large-scale text analysis cheaper and better at the same time.
Catalog record
| CREATOR | Lindsey Linxi Wei 1† , Shreya Shankar 2† , Sepanta Zeighami 2 , Yeounoh Chung 3 , Fatma Ozcan 3 , Aditya G. Parameswaran 2 |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 123 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-02289