TabHGR

Code and datasets of Zero-Shot Tabular Learning via Hypergraph Reasoning with LLMs (EMNLP 2026).

Tabular Datasets

Experiments can be run on nine tabular datasets: adult, bank, blood, car, credit-g, diabetes, heart, communities, and myocardial.

Installation

Install the required packages:

python -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install -r requirements.txt

Generating Weak Labels and Groups (Optional)

The generated weak labels and groups are already included. To generate them again:

export OPENAI_API_KEY="your-api-key"
python generate_weak_labels.py --dataset diabetes
python generate_groups.py --dataset diabetes

Running TabHGR and TabHGR+

Run TabHGR and TabHGR+ on a dataset:

python main.py --dataset diabetes

To run specific trials:

python main.py --dataset diabetes --trials 0
python main.py --dataset diabetes --trials 1,2,3,4

If memory is limited, run one trial at a time:

OPENBLAS_NUM_THREADS=1 python main.py --dataset adult --trials 0

To view all available options:

python main.py --help