Credibility track · Astronomy · Moderated journal
Research Notes of the AAS
The American Astronomical Society's journal for short, timely results. It is moderated by an editor for appropriateness and format rather than peer-reviewed, publishes within days, is free, and gives every note a DOI and an ADS listing. This is the first publication for this project.
Why this is the near-certain one: RNAAS checks that a note is appropriate astronomy and properly formatted. It doesn't ask for novelty or completeness, and accepted notes are typically online within 72 hours. It still carries the AAS name, a DOI and an ADS entry, which is real, citable proof that the work exists.
Honest caveat: moderation is not a rubber stamp. The editor can decline a note that reads as out of scope or as an unfounded theory. This note is a controlled experiment with public code and conservative claims, which is the kind RNAAS publishes.
Venue facts
| Length | ≤ 1,500 words including title, headers, captions and references. The required abstract is ≤ 150 words. Table contents don't count. One figure or one table, not both. |
|---|---|
| Review | Not peer-reviewed. Moderated by an editor for appropriateness and format, published as submitted with no copy-editing, typically within 72 hours of acceptance. |
| Cost and licence | No author charges. Gold open access, CC BY 4.0. |
| Indexing | DOI, indexed in NASA ADS. arXiv accepts RNAAS notes. |
| Format | AASTeX v7 (preferred), Word or Overleaf. No line numbers. Single authors write in the first person singular. |
| Submit at | aas.msubmit.net (Editorial Office). |
| Sources | RNAAS preparation guidelines · Research Notes of the AAS |
The note
Keeping a “None of the Above” Hypothesis Matters More Than Planning With It: A Controlled Test for Transient Follow-Up
Information-driven follow-up planners for astronomical transients usually assume the event belongs to one of a fixed set of classes. We test, in a controlled toy problem, what happens when it does not. Across simulated fading transients ( seeds), a closed-world expected-information-gain planner was confidently wrong on of events from a withheld class. Adding an explicit unmodeled hypothesis h₀, represented by a Gaussian process, to the posterior removes most of this failure. Planning with h₀ helped only modestly: the open-world planner flagged of withheld events, against for the closed-world planner's measurements rescored with h₀. Randomly timed measurements rescored the same way flagged . Information-gain planning, even with h₀, chose measurements that exposed model inadequacy less well than random timing, so inadequacy detection has to be designed into the planning objective explicitly.
Read the compiled draft (PDF) →
Results at a glance
| Measurements chosen by | Scored with | Confidently wrong (withheld) | Flagged h₀ (withheld) | AUROC of P(h₀) |
|---|---|---|---|---|
| Random | closed-world | — | ||
| Random | open-world (post hoc) | |||
| Closed-world EIG | closed-world | — | ||
| Closed-world EIG | open-world (post hoc) | |||
| Open-world EIG (P) | open-world |
Source: research/results/rnaas_note/numbers.json, generated by research/figures/rnaas_note.py from research/results/phase1_full/ (seeds ).
Submission steps
- Run the controlled experiment at full scale (3 seeds × 1,000 events) and generate every number from result files.
- Draft in AASTeX v7 within the limits: ≤150-word abstract, one table.
- Replace the author placeholders in
research/papers/rnaas/note.tex: name, ORCID, affiliation, email, any co-authors. Single-author notes should use “I”, not “we”. - Add a LICENSE file to the repository, tag a release, and mint a Zenodo DOI. Cite it in Code availability.
- Rebuild:
python -m research.figures.rnaas_note, thenbash research/papers/rnaas/build.sh. Check the editorial word count. - Submit at aas.msubmit.net. After acceptance, post to arXiv (astro-ph.IM) and add the DOI to this site.