ADO: Automatic Data Optimization for Inputs in LLM Prompts

Sam Lin*, Wenyue Hua*, Lingyao Li, Zhenting Wang, Yongfeng Zhang · 2025

Read ADO: Automatic Data Optimization for Inputs in LLM Prompts by Sam Lin*, Wenyue Hua*, Lingyao Li, Zhenting Wang, Yongfeng Zhang (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-2502-11436.txt.

Description

Sam Lin and colleagues argue that the data inside a prompt deserves the same care as the instructions around it. Most prompt engineering polishes wording or adds worked examples, leaving the raw input untouched. Writing in twenty twenty-five, the team instead optimizes the input itself in two stages. Content engineering fills missing values, drops irrelevant fields, and infers extra profile details from what is present. Structural reformulation then reshapes the cleaned content into the layout a given model reads best, since models are sensitive to format.

Across several tasks, these input fixes raised model performance by clear margins, according to the authors tests. Gains came without changing the model or the task instructions. The paper opens a less explored branch of prompt work, treating data quality and presentation as first-class levers.

Catalog record

CREATORSam Lin*, Wenyue Hua*, Lingyao Li, Zhenting Wang, Yongfeng Zhang
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE40 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-11436

Browse the full library · About the Tower of Babel