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..

Quantum Visual Fields with Neural Amplitude Encoding

Authors: Shuteng Wang, Christian Theobalt, Vladislav Golyanik

NeurIPS 2025 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on a quantum hardware simulator demonstrate that QVF outperforms existing quantum approach and competes widely used classical foundational baselines in terms of visual representation accuracy across various metrics and model characteristics.
Researcher Affiliation Academia MPI for Informatics, SIC, Germany
Pseudocode Yes We summarise the QVF training protocol in Algorithm 1 in the Appendix.
Open Source Code No all necessary algorithmic details are provided in the main text to re-produce the results. and the source code will be made available when the paper is released.
Open Datasets Yes We use 1) images from the CIFAR-10 dataset [24] and high-resolution images with rich spectral details [15], and 2) 3D shapes from the Shape Net [10] dataset.
Dataset Splits No The paper mentions using CIFAR-10 and ShapeNet datasets, and evaluates performance on '50 different images' and 'three shapes from Shape Net' with '100k spatial points per shape'. However, it does not explicitly provide percentages, absolute sample counts, or citations for specific training/test/validation splits of these datasets.
Hardware Specification Yes We empirically evaluate the model on a noiseless high-end simulator: default.qubit.torch, provided by Penny Lane [6] with an A100 GPU.
Software Dependencies No The paper mentions 'default.qubit.torch, provided by Penny Lane [6]' for the simulator, and 'Adam optimisation [23]'. However, it does not provide specific version numbers for Penny Lane or default.qubit.torch.
Experiment Setup Yes We employ Adam optimisation [23] with an initial learning rate of η = 10 3, subject to a learning rate scheduler that triggers upon plateauing with a window size of 50 epochs (scaling η by 0.9). The number of epochs is set to 5k, and γ = 10 3 in Eq. (12).