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..
Robust Hyperbolic Learning with Curvature-Aware Optimization
Authors: Ahmad Bdeir, Johannes Burchert, Lars Schmidt-Thieme, Niels Landwehr
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our approach demonstrates consistent performance improvements across Computer Vision, EEG classification, and hierarchical metric learning tasks while greatly reducing runtime. Our code is publicly available at https://github.com/inboxedshoe/Robust-Hyperbolic-Learning. ... We empirically show the effectiveness of our proposed methods in five domains, hierarchical metric learning, EEG classification, graph embedding, image classification, and image generation to show the effectiveness of our optimizer in different problem settings. We improve performance in all domains with a significant computational speed-up. |
| Researcher Affiliation | Academia | Ahmad Bdeir Department of Data Science University of Hildesheim Hildesheim, Germany EMAIL Johannes Burchert ISMLL University of Hildesheim Hildesheim, Germany EMAIL Lars Schmidt-Thieme ISMLL University of Hildesheim Hildesheim, Germany EMAIL Niels Landwehr Department of Data Science University of Hildesheim Hildesheim, Germany EMAIL |
| Pseudocode | Yes | Algorithm 1 Tangent Based Manifold Mapping ... Algorithm 2 Riemannian Adam (RAdam) and Riemannian Adam W (RAdam W) |
| Open Source Code | Yes | Our code is publicly available at https://github.com/inboxedshoe/Robust-Hyperbolic-Learning |
| Open Datasets | Yes | We use only publicly available datasets and have attached our code in the supplemental material. |
| Dataset Splits | Yes | We do specify the dataset splits and, where applicable, use the same ones already established in the literature. |
| Hardware Specification | Yes | All experiments were conducted on a single NVIDIA RTX 4090 GPU and an AMD EPYC 7543 CPU. |
| Software Dependencies | No | The paper mentions software like PyTorch (implied by references to Riemannian optimization libraries and common deep learning frameworks), but does not provide specific version numbers for any software libraries or dependencies. The reproducibility checklist states: "We also provide a README in our code with all hyperparameters that can be optimized and state which optimizer is being used.", but this doesn't explicitly list software versions in the paper itself. |
| Experiment Setup | Yes | We follow the experimental setup in Kim et al. [19]. ... We keep the same hyperparameters defined in Chen et al. [6] for all training settings. ... We also provide a README in our code with all hyperparameters that can be optimized and state which optimizer is being used. |