This repository contains model outputs from our study on neural network calibration across various architectures and datasets.

Available Datasets
==================

- ImageNet
- ImageNet-V2
- ImageNet-A
- ImageNet-C

Available Models
================ 

- ConvNeXt
- EVA
- BEiT
- ViT
- Swin Transformer
- ResNet

File Contents
=============

Each .safetensor file includes:
- Model logits and ground truth labels for the validation set (for the distribution-shifted datasets)
- Model logits and ground truth labels for the test set

File Naming Convention
======================
[model_name]_output.safetensor, where model_name is the string identifier from the timm library.

Important Note
==============
Due to the size of the original data (>100 GB), it is not possible to upload the complete dataset in an anonymized way for submission.
We have therefore only uploaded a very small subset of the data set here. If the paper is accepted, the data will be uploaded in full to zenodo.