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..
Low Rank Gradients and Where to Find Them
Authors: Rishi Sonthalia, Michael Murray, Guido F. Montufar
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on synthetic and real data corroborate our theoretical predictions. |
| Researcher Affiliation | Academia | Rishi Sonthalia Department of Mathematics Boston College EMAIL Michael Murray Department of Mathematics Bath University EMAIL Guido Montúfar Mathematics and Statistics & Data Science University of California, Los Angeles EMAIL |
| Pseudocode | No | The paper describes methods and derivations in textual and mathematical forms but does not contain any clearly labeled 'Pseudocode' or 'Algorithm' blocks. |
| Open Source Code | Yes | All code is available at the anonymous Github repository: https://github.com/rsonthal/Low-Rank-Gradient |
| Open Datasets | Yes | Experiments on synthetic and real data (MNIST, CIFAR-10 embeddings). |
| Dataset Splits | No | The paper mentions using specific numbers of samples for real datasets (e.g., '1000 centered and flattened MNIST images', 'n = 1000 CIFAR-10 training images') but does not explicitly specify how these samples are split into training, validation, or test sets for their experiments, nor does it refer to standard splits with citations for the experimental setup. |
| Hardware Specification | Yes | Hardware: All experiments were run on Google Colab using an A100. |
| Software Dependencies | No | The paper mentions using components like 'resnet18' and 'transforms.Resize' in the context of CIFAR-10 processing, which implies the use of the PyTorch library. However, it does not provide specific version numbers for PyTorch or any other software dependencies. |
| Experiment Setup | Yes | Fixed parameters: ν = 1/8, α = 5/9, n = 750, d = 1000, and m = 1250. The following parameters are constant across all experiments: α = 0, γm = 1/m (NTK) n = 750, d = 1000, m = 1250. We use a step size of γm^-1. Additionally, after each iteration, we re-normalize the rows of W to have unit norm. |