A Mechanistic Account of Dedifferentiation in the Aging Visual System

This record contains the simulation outputs behind every figure and table of A Mechanistic Account of Dedifferentiation in the Aging Visual System. Pretrained deep neural networks (CORnet-RT, CORnet-S, AlexNet, VGG16) were damaged by connection severing, neuron deletion or weight noise, at increasing intensities and in different model areas, including damage to V1 that depends on eccentricity. The record includes: the representational dissimilarity matrices (RDMs) of the category-selective readout units for every damage condition and level; the category differentiation tables computed from them; the unit selectivity tables; the IT-to-V1 gradient maps and eccentricity profiles; the statistics behind Tables S1 and S2. Each simulation is identified by its replicate number, and the stimuli in each RDM by name and category. Start with dedifferentiation_core.zip (~0.4 GB), which is enough to redraw every figure and table. The 26 RDM zips (~27 GB in total) are needed only to rebuild the differentiation tables. README.md maps each file to the manuscript figures and tables, and MANIFEST.tsv lists the size and SHA-256 hash of every zip. The code is at https://github.com/AkillesU/dedifferentiation-models; its download_data.py fetches and unpacks this record. The experimental stimuli and human data of Haupt et al. (2025) are not redistributed here. The stimuli can be downloaded from their OSF project (https://osf.io/xeukw/), and the human reference data are in the code repository.