Status: skeleton — technical report intended for Zenodo. Not peer-reviewed.
How do CEUR-WS papers declare the use of Generative AI, and how well do the checks at the creation points detect a missing or generic Declaration on Generative AI? This repository holds the report source and, as the work proceeds, the scripts, sample, verdicts and ground truth it reports on.
- Experiment design: CEUR-WS Declaration on Generative AI Check
- Creation points: semantify discussion #40
- PDF: main.pdf (built by CI)
The PDF is rebuilt on every push to main via GitHub Actions
(.github/workflows/build.yml).
Official CEUR-WS ceurart 1-column class.
Work is issue driven with agentic AI support: main issues are dictated by the authors, derived sub-issues and experiment runs may be done by agents and are marked per the provenance notation below.
Every issue comment carries its origin as a prefix:
| marker | meaning |
|---|---|
| ✍️🤖 Agent/Name: | generated by the agent |
| 🗣️ Author-Name → 🤖 Agent/Name: | dictated by the human, transcribed and posted by the agent |
| 👁️🧑 Author-Name ✓ | the human read the agent result and confirmed it |
| 👁️🧑 Author-Name ❓ | the human sees an open question |
| 👁️🧑 Author-Name ✗ | the human rejects the result |
Only ✓ counts as confirmed. PROVENANCE.md lists every agent contribution to the repository; the Declaration on Generative AI is finalized by the authors from it.
See AGENTS.md.
See data/README.md.
Report text and data: CC BY 4.0. Code: Apache-2.0.