Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Large Language Models Can Automatically Engineer Features for Few-Shot Tabular Learning

About

Large Language Models (LLMs), with their remarkable ability to tackle challenging and unseen reasoning problems, hold immense potential for tabular learning, that is vital for many real-world applications. In this paper, we propose a novel in-context learning framework, FeatLLM, which employs LLMs as feature engineers to produce an input data set that is optimally suited for tabular predictions. The generated features are used to infer class likelihood with a simple downstream machine learning model, such as linear regression and yields high performance few-shot learning. The proposed FeatLLM framework only uses this simple predictive model with the discovered features at inference time. Compared to existing LLM-based approaches, FeatLLM eliminates the need to send queries to the LLM for each sample at inference time. Moreover, it merely requires API-level access to LLMs, and overcomes prompt size limitations. As demonstrated across numerous tabular datasets from a wide range of domains, FeatLLM generates high-quality rules, significantly (10% on average) outperforming alternatives such as TabLLM and STUNT.

Sungwon Han, Jinsung Yoon, Sercan O Arik, Tomas Pfister• 2024

Related benchmarks

TaskDatasetResultRank
ClassificationAdult
Accuracy84.2
86
ClassificationDiabetes
Accuracy84.9
80
Classificationvehicle
Accuracy74.4
65
ClassificationCredit
ROCAUC74.3
63
ClassificationHeart
Accuracy86.5
59
Clinical predictionMIMIC-III
AUROC90.4
59
ClassificationBank--
48
ClassificationCAR
Accuracy80.8
47
Multiclass ClassificationCMC
Accuracy47.9
41
ClassificationBreastW
Accuracy96.7
29
Showing 10 of 71 rows
...

Other info

Follow for update