Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..

Task-Agnostic Machine-Learning-Assisted Inference

Authors: Jiacheng Miao, Qiongshi Lu

NeurIPS 2024 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Through extensive experiments, we showcase our method s validity, versatility, and superiority compared to existing approaches.
Researcher Affiliation Academia Jiacheng Miao University of Wisconsin-Madison EMAIL Qiongshi Lu University of Wisconsin-Madison EMAIL
Pseudocode Yes Algorithm 1 PSPS for ML-assisted inference
Open Source Code Yes Our software is available at https://github.com/qlu-lab/psps.
Open Datasets Yes We used data from the UK Biobank [13], which includes 36,971 labeled and 319,548 unlabeled samples with 9,450,880 genetic variants after quality control.
Dataset Splits Yes Prediction in the labeled sample was implemented through cross-validation to avoid overfitting. The implementation detail is deferred to Appendix D. We select the predictive variables and train the Soft Impute model using 90% of the labeled data. We then perform predictions on the remaining 10% in each fold and repeat this process 10 times across all folds.
Hardware Specification Yes All our simulation is run in R with version 4.2.1 (2022-06-23) in a Mac Book Air with an M1 chip.
Software Dependencies Yes All our simulation is run in R with version 4.2.1 (2022-06-23) in a Mac Book Air with an M1 chip.
Experiment Setup Yes A pre-trained random forest with 500 trees to grow is obtained from hold-out data. We bootstrap the labeled data for 200 times for covariance estimation. All simulations are repeated 1000 times.