LLM-enhanced Score Function Evolution for Causal Structure Learning

Authors: Zidong Wang, Fei Liu, Qi Feng, Qingfu Zhang, Xiaoguang Gao

IJCAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental evaluations, conducted on discrete, continuous, and real datasets, demonstrate the high stability, generality and effectiveness of L-SFE. (Abstract)
Researcher Affiliation Academia 1Department of Computer Science, City University of Hong Kong, Hong Kong 2School of Electronic and Information, Northwestern Polytechnical University, Xi an, China EMAIL, EMAIL, EMAIL, EMAIL.
Pseudocode Yes To provide a clear understanding of L-SFE, we outline the pseudo-code in Alg. 1 in Supplementary material 1.2.
Open Source Code Yes Code is avaliable on https://github.com/wzd2502/L-SFE
Open Datasets Yes synthetic datasets generated from pytetrad are employed for training and testing. For the discrete dataset... For the continuous datasets... We present a case study using data from the real world COVID-19 pandemic in the UK... https://bayesian-ai.eecs.qmul.ac.uk/bayesys/
Dataset Splits No The paper describes generating synthetic datasets for training and testing and mentions using a real-world COVID-19 dataset with 866 samples, but does not specify explicit training/test/validation splits (e.g., percentages or counts) for a single dataset.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU models, CPU types) used for running the experiments.
Software Dependencies No The paper mentions "GPT-4o mini is utilized for score function discovery in L-SFE" and "synthetic datasets generated from pytetrad are employed", but does not provide specific version numbers for these software components.
Experiment Setup Yes For the discrete dataset, L-SFE is trained on ten Random Graphs with n = 30... Each test is repeated 10 times with m = 5000. ... The GLS used for training is HC with a tabu search.