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..
Neural Collapse is Globally Optimal in Deep Regularized ResNets and Transformers
Authors: Peter Súkeník, Christoph H. Lampert, Marco Mondelli
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our theoretical results are supported by experiments on computer vision and language datasets showing that, as the depth grows, neural collapse indeed becomes more prominent. 5 Experimental results |
| Researcher Affiliation | Academia | Peter Súkeník Institute of Science and Technology (ISTA) Austria EMAIL Christoph H. Lampert Institute of Science and Technology (ISTA) Austria EMAIL Marco Mondelli Institute of Science and Technology (ISTA) Austria EMAIL |
| Pseudocode | No | The paper describes methods mathematically and textually but does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | No | Question: Does the paper provide open access to the data and code, with sufficient instructions to faithfully reproduce the main experimental results, as described in supplemental material? Answer: [No] Justification: We do not find it necessary to release the code. Our experiments concern the training of rather standard architectures and they can be readily reproduced without needing to upload the code. |
| Open Datasets | Yes | we train Res Nets and transformers on MNIST [29], CIFAR10 [31] and IMDB [21] |
| Dataset Splits | No | The paper mentions training on MNIST, CIFAR10, and IMDB datasets but does not explicitly state the training, validation, or test splits used, nor does it refer to specific standard splits. |
| Hardware Specification | No | Question: For each experiment, does the paper provide sufficient information on the computer resources (type of compute workers, memory, time of execution) needed to reproduce the experiments? Answer: [No] Justification: Our experiments require only modest computational resources. |
| Software Dependencies | No | The paper does not specify any software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow, CUDA versions) needed to replicate the experiments. |
| Experiment Setup | Yes | The hidden dimension is 64, the learning rate 0.005 for vision and 0.001 for language and the (constant) regularization 0.005 for architectures having one linear layer per block and 0.005/L for architectures having two linear layers per block. Each setting is trained for 5 different random seeds for 5000 epochs on CE loss |