# Third-party notices

This bundle contains generated predictions and processed data from several
upstream resources. The terms listed below apply to their respective
components; they are not replaced by a single bundle-wide license.

## UNI2-H

The distributed GHIST+ checkpoints do **not** contain UNI2-H tensors. Users
obtain the official checkpoint directly from:

- Repository: https://huggingface.co/MahmoodLab/UNI2-h
- File: pytorch_model.bin

Access and use are governed by the upstream UNI2-H terms, including the
recorded CC BY-NC-ND 4.0 and noncommercial-use conditions. Reconstructed
checkpoints that incorporate UNI2-H remain subject to those terms.

## Public source datasets

The evaluation inputs are processed derivatives of public 10x Genomics Xenium
datasets, including:

- Atera WTA FFPE Human Breast Cancer
- Xenium FFPE Human Breast Cancer Sample 1, Replicates 1, 2, and 4
- Xenium Prime 5K Pan-Tissue and Pathways Panel, FFPE Human Breast Cancer

Source: https://www.10xgenomics.com/datasets

The source datasets are distributed by 10x Genomics under their stated terms
and CC BY 4.0 notices. Files in this bundle are modified/processed versions,
not unmodified source archives.

## SpatialEx

SpatialEx comparison outputs were generated with code from:

- Repository: https://github.com/KEAML-JLU/SpatialEx
- Upstream code license: MIT

Outputs generated using UNI2-H are also subject to the applicable UNI2-H
noncommercial terms.

## PASTA

PASTA comparison outputs were generated from:

- Repository: https://github.com/rx-li/PASTA
- Archived release: https://doi.org/10.5281/zenodo.17336118

Refer to the upstream repository, release record, and publication for the
applicable terms and citation.

## Phoenix

Phoenix comparison outputs were generated from:

- Repository: https://github.com/peng-lab/phoenix
- Model repository: https://huggingface.co/peng-lab/phoenix

The Phoenix code terms are PolyForm Noncommercial 1.0.0 and the model
weights are CC BY-NC 4.0. The released files are generated predictions, not
Phoenix code or model weights.

## GHIST+ source code

GHIST+ source code and its license are provided separately at:
https://github.com/SydneyBioX/GHIST_plus