distribution-shift
gpt-4v-distribution-shift
License
This repository is licensed under the MIT License.
Description
This Hugging Face repository hosts the random case dataset utilized in our research project, detailed in the GitHub repository gpt-4v-distribution-shift.
These datasets are crucial for evaluating the performance of multimodal foundation models under various distribution shift scenarios.
Using the Dataset
For detailed instructions on how to use this dataset to reproduce the results presented… See the full description on the dataset page: https://huggingface.co/datasets/jameszhou-gl/gpt-4v-distribution-shift.layout_distribution_shiftmlff_distribution_shiftsThis repository provides the benchmark datasets used in the paper "Understanding and Mitigating Distribution Shifts For Machine Learning Force Fields" (https://arxiv.org/abs/2503.08674). Check out the project page: https://tkreiman.github.io/projects/mlff_distribution_shifts/
tree-distribution-shift
Tree Distribution Shift — Satellite Tree Detection (COCO + HF Datasets)
This is a dataset containing ~30K COCO tree crown annotated satellite image tiles of 400x400 px dimensions. These annotations come from all states in India and California in the United States.
This dataset is organized as configs (distribution shift benchmarks). Each config provides three splits:
train — 90% of the in-distribution pool
id_test — 10% held-out from the in-distribution pool
ood_test —… See the full description on the dataset page: https://huggingface.co/datasets/aadityabuilds/tree-distribution-shift.