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 [1].

TSP: A Two-Sided Smoothed Primal-Dual Method for Nonconvex Bilevel Optimization

Authors: Songtao Lu

ICML 2025 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Numerical experiments underscore the superiority of our proposed algorithm over existing penalty-based methods in terms of both the convergence rate and the test accuracy.
Researcher Affiliation Academia 1Department of Computer Science and Engineering and Shun Hing Institute of Advanced Engineering, The Chinese University of Hong Kong, Hong Kong. Correspondence to: Songtao Lu <EMAIL>.
Pseudocode Yes Algorithm 1 Single-loop stochastic Two-sided Smoothed Primal-dual (TSP) method for bilevel optimization
Open Source Code No The paper does not explicitly provide a link to open-source code or state that code will be released, either in the main text or supplementary materials.
Open Datasets Yes Specifically, we use the MNIST dataset, splitting it into three parts: 5,000 training samples, 5,000 validation samples, and 10,000 test samples.
Dataset Splits Yes Specifically, we use the MNIST dataset, splitting it into three parts: 5,000 training samples, 5,000 validation samples, and 10,000 test samples. Additionally, 50% of the training data samples are randomly assigned incorrect labels as polluted data.
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 does not specify version numbers for any software components, libraries, or programming languages used in the experiments.
Experiment Setup Yes For TSP, we further choose the step-sizes for updating the dual variable λ as 0.01 and set p = 1. ... In the numerical experiments, the selected learning rates for all algorithms are 0.01 for x, 0.05 for y, and 0.06 for z. The dual variable learning rate for TSP is 0.5, selected from {1, 0.5, 0.1, 0.01}. ... The gradients used in all implemented algorithms are stochastic, with a batch size of 32.