# Attribution and licenses:

The original project and new analysis scripts are distributed under the included Apache License 2.0. This notice does not replace upstream terms for third-party software, pretrained weights or benchmark data.

## GridFM graphkit:

- Repository: https://github.com/gridfm/gridfm-graphkit
- Recorded snapshot: `4e7fbcf6aa33c09b5ec9f15e5952dbbff004b74e` (`genco-paper-repro`).
- License and modification notice: `code/src/genco_resilience/_vendor/gridfm/LICENSE` and `NOTICE.md` in that directory.
- Modifications described there include namespace isolation, native additive scatter, explicit layer budgeting, diagnostic controls and the OPF task registry. The pretrained checkpoint tensors remain unchanged.

## GridFM DataKit:

- Repository: https://github.com/gridfm/gridfm-datakit
- Recorded snapshot: `c344b65279a840ba6f49a56878c719813b6ac9c0`.
- The vendored power-balance source is unmodified; its hashes, license and notice are under `code/src/genco_resilience/_vendor/datakit/`.

## Pretrained checkpoints:

`code/workload_checkpoint_lock.json` specifies immutable Hugging Face revisions and file hashes. Checkpoints are fetched from their original repositories. Original model authorship and model-card terms apply. PF4 changes computation depth using the same PF weights. This release supplies no newly trained checkpoint.

## Network cases:

The retained `pglib_opf_case14_ieee.m`, `pglib_opf_case30_ieee.m` and `pglib_opf_case118_ieee.m` include their original provenance and Creative Commons Attribution 4.0 notices in the source headers. Preserve those headers and citations when redistributing the cases. The report cites MATPOWER and the original case sources remain in the files. The Apache label for project software does not relicense these cases.

## IAS LaTeX template:

- Repository: https://github.com/IAS-Uni-Siegen/answer_letter
- `report/reviewresponse.sty` is retained from that template and used by `report/report.tex`.
- Copyright (c) 2025 Chair of Interconnected Automation Systems (IAS).
- Its MIT license is included as `report/template-LICENSE.txt`.

## Runtime dependencies:

PyTorch, PyG, CasADi/IPOPT, PYPOWER, Julia, ExaModels, MadNLP and their transitive dependencies retain their respective licenses. They are referenced through installation manifests, not bundled as runtime binaries. The Julia manifest records the resolved dependency graph. Direct observed Python versions are recorded separately.

## Development assistance:

The study was directed by Bikash Sah. ChatGPT/Codex assisted with software development, analysis and submission preparation under the author's direction. The research authorship and responsibility statement is provided in the report.
