Newest papers

When Do LLMs Help with Feature Selection? A Systematic Empirical Study

Fateme Memar, Dongjie Wang, Arvin Agah · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.28556 · cited by 0 works (OpenAlex) · open-access PDF

French

Les grands modèles de langue sélective (LLM) peuvent tirer parti des noms de caractéristiques sémantiques aux côtés des statistiques de synthèse pour la sélection des caractéristiques de table, mais il reste incertain quand ces informations améliorent la sélection des sous-catégories par rapport aux méthodes statistiques établies. Nous enquêtons sur cette question dans un cadre de 18 ensembles de données comprenant 10 ensembles de données pour la formation ROC-AUC relative (GRPO) et la régression R2 séparément, 2 pour la validation des points de contrôle et 6 ensembles de données d'évaluation détenus pour la comparaison. Nous comparons cinq lignes de base classiques, les sélecteurs de classe congelés Qwen2.5-7B et Llama-3.3-70B, et deux mod

Machine translation (Meta M2M100); the English original follows.

Abstract

Large language models (LLMs) can leverage semantic feature names alongside summary statistics for tabular feature selection, yet it remains unclear when such information improves subset selection relative to established statistical methods. We investigate this question within an 18-dataset framework comprising 10 datasets for Group Relative Policy Optimization (GRPO) training, 2 for checkpoint validation, and 6 held-out evaluation datasets. We compare five classical baselines, frozen Qwen2.5-7B and Llama-3.3-70B selectors, and two GRPO-adapted Qwen2.5-7B models, reporting classification ROC-AUC and regression R^2 separately and using dataset-wise ranks for cross-task comparison. Across five seeded evaluation partitions, classical selectors generally achieve higher predictive performance than the tested 7B LLM/RL selectors, including under a pre-specified ~33% feature budget with strict train-only classical fitting, while the 7B selectors retain compact subsets. A cardinality-matched semantic sensitivity experiment shows that, for Zero-shot Qwen-7B, informative feature names improve regression performance over anonymized and shuffled alternatives (+0.1964 and +0.1606 mean R^2), whereas aggregate classification differences are smaller (+0.0159 and +0.0084 ROC-AUC) and dataset-dependent. These results characterize when semantic information benefits LLM-based feature selection and show that classical selectors remain strong baselines for predictive tabular feature selection.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Feature selection, Leverage (statistics), Feature (linguistics), Computer science, Artificial intelligence, Regression

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange