Saved in:
Bibliographic Details
Main Authors: Cosma, Adrian, Szehr, Oleg, Kletz, David, Antonucci, Alessandro, Pelletier, Olivier
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2601.13922
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914266255196160
author Cosma, Adrian
Szehr, Oleg
Kletz, David
Antonucci, Alessandro
Pelletier, Olivier
author_facet Cosma, Adrian
Szehr, Oleg
Kletz, David
Antonucci, Alessandro
Pelletier, Olivier
contents Feature extraction from unstructured text is a critical step in many downstream classification pipelines, yet current approaches largely rely on hand-crafted prompts or fixed feature schemas. We formulate feature discovery as a dataset-level prompt optimization problem: given a labelled text corpus, the goal is to induce a global set of interpretable and discriminative feature definitions whose realizations optimize a downstream supervised learning objective. To this end, we propose a multi-agent prompt optimization framework in which language-model agents jointly propose feature definitions, extract feature values, and evaluate feature quality using dataset-level performance and interpretability feedback. Instruction prompts are iteratively refined based on this structured feedback, enabling optimization over prompts that induce shared feature sets rather than per-example predictions. This formulation departs from prior prompt optimization methods that rely on per-sample supervision and provides a principled mechanism for automatic feature discovery from unstructured text.
format Preprint
id arxiv_https___arxiv_org_abs_2601_13922
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Automatic Prompt Optimization for Dataset-Level Feature Discovery
Cosma, Adrian
Szehr, Oleg
Kletz, David
Antonucci, Alessandro
Pelletier, Olivier
Computation and Language
Feature extraction from unstructured text is a critical step in many downstream classification pipelines, yet current approaches largely rely on hand-crafted prompts or fixed feature schemas. We formulate feature discovery as a dataset-level prompt optimization problem: given a labelled text corpus, the goal is to induce a global set of interpretable and discriminative feature definitions whose realizations optimize a downstream supervised learning objective. To this end, we propose a multi-agent prompt optimization framework in which language-model agents jointly propose feature definitions, extract feature values, and evaluate feature quality using dataset-level performance and interpretability feedback. Instruction prompts are iteratively refined based on this structured feedback, enabling optimization over prompts that induce shared feature sets rather than per-example predictions. This formulation departs from prior prompt optimization methods that rely on per-sample supervision and provides a principled mechanism for automatic feature discovery from unstructured text.
title Automatic Prompt Optimization for Dataset-Level Feature Discovery
topic Computation and Language
url https://arxiv.org/abs/2601.13922