Citation, credit and licenses¶

K GraphFuse is created and maintained by Utkarsh Sinha. The project requests visible attribution when it supports a product, project or paper.

Suggested public product credit:

Powered by K GraphFuse by Utkarsh Sinha

Link that text to the repository. For an improved derivative, a useful description is “Built on K GraphFuse by Utkarsh Sinha,” followed by your changes and your own authorship. Place credit in a readable acknowledgment, documentation or product technology section.

Cite this release¶

GitHub's Cite this repository control uses CITATION.cff. A reusable BibTeX entry is:

@software{sinha2026kgraphfuse,
  author  = {Sinha, Utkarsh},
  title   = {K GraphFuse: sparse graph polynomial plans and CUDA execution},
  year    = {2026},
  version = {0.1.0},
  url     = {https://github.com/Utkarsh-Sinha0/k-graphfuse}
}

Add the actual release/commit used in reproducible studies. Raw experiments retain the earlier Sinha GraphFuse name for traceability; this is the same numerical core under a new public brand.

Cite the technical manuscript¶

The research manuscript is a technical preprint draft first posted 10 October 2026 and revised 11 October 2026. It has not been peer-reviewed or submitted as a journal publication, and has no assigned DOI. Cite the actual revision used:

@misc{sinha2026similaritygraphfuse,
  author = {Sinha, Utkarsh},
  title  = {{K GraphFuse}: Similarity-Aware Sparse Polynomial Evaluation
            with Selective GPU Compensation},
  year   = {2026},
  note   = {Technical manuscript, preprint draft, revised 11 October 2026},
  url    = {https://utkarsh-sinha0.github.io/k-graphfuse/research/paper/}
}

The PDF and editable LaTeX source accompany the web manuscript. Use the software citation as well when reproducing its code.

License scope¶

New namespace, documentation, examples and repository infrastructure use Apache License 2.0. The historical numerical modules, original tests and benchmark fixtures retain the original MIT grant in LICENSES/MIT.txt. The distribution carries both notices; the SPDX expression is Apache-2.0 AND MIT.

Preserve the applicable copyright, license and NOTICE information when redistributing. Apache 2.0 does not make the requested frontend credit mandatory. This project does not add a restrictive attribution clause or source-sharing obligation. Independent reuse of mathematical ideas or principles is outside a software-copyright claim; the US Copyright Office explains that distinction.

Scientific and external attribution¶

The derivation attributes the complete-inequality manuscript, Crouzeix–Palencia and derivative prior art. Graphalytics, SNAP, NVIDIA, ArcadeDB, PyTorch and other external materials retain their owners' terms. Data and vendor binaries are acquired separately. Their names indicate provenance and comparison, not affiliation or endorsement. See mathematics and benchmark results.

The revised manuscript separates the supplied OpenAI theorem from K GraphFuse's computational application and measured attribution. Cite the underlying theorem and the software when using this numerical approach. No original-theorem authorship or worldwide-first algorithm claim is implied.