Notice: This website is continuously updated as new conference proceedings and journal volumes are published. Consequently, the figures presented here may differ from those reported in our publications.
A brief overview of our methodology is provided on the Methods page, along with references to publications documenting our approach and links to the code and data used in each study.
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.
Large Stepsizes Accelerate Gradient Descent for Regularized Logistic Regression
Authors: Jingfeng Wu, Pierre Marion, Peter L Bartlett
NeurIPS 2025 | Venue PDF | LLM Run Details | Input Tokens: 34,237 Total number of tokens sent to the LLM as input for this paper's analysis. | Output Tokens: 3,591 Total number of tokens produced by the LLM (including reasoning/thinking tokens) for this paper's analysis.
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Simulations. Our results are illustrated in Figure 2 by running GD for ℓ2-regularized logistic regression on a toy two-dimensional separable dataset. ... Additional plots are given in Appendix E. |
| Researcher Affiliation | Collaboration | Jingfeng Wu UC Berkeley EMAIL Pierre Marion Inria, DI ENS, PSL University EMAIL Peter L. Bartlett UC Berkeley & Google Deep Mind EMAIL |
| Pseudocode | Yes | wt+1 := wt η e L(wt), t 0, w0 H, (GD) |
| Open Source Code | Yes | Our code is available at https://github.com/Pierre Marion23/large-stepsize-regularized-logistic. |
| Open Datasets | Yes | The dataset is composed on two datapoints x1 = (γ, 1) and x2 = (γ, 2) for γ = 0.2. |
| Dataset Splits | No | The paper uses a synthetic dataset with only two datapoints, x1 and x2, as defined in Appendix E. In such a context, traditional training, validation, and test splits are not applicable or provided. |
| Hardware Specification | No | The code was implemented in JAX (Bradbury et al., 2018) and takes a few seconds to run on a consumer laptop. |
| Software Dependencies | No | The code was implemented in JAX (Bradbury et al., 2018)... |
| Experiment Setup | Yes | The dataset is composed on two datapoints x1 = (γ, 1) and x2 = (γ, 2) for γ = 0.2. We run GD on the regularized logistic regression for λ = 2 12, a logarithmic range of stepsizes from 21 to 213, and 213 steps. |