Skip to content
Apart Research
Apart Research
Contact us
Sprint projectJul 27, 2026Princeton NJ

A Multi-Principal Organism Specification with Matched Controls and a Negative-Control Principal

Aayush Gandhi · Team Secret_Team

Submitted to Secret Loyalties Hackathon. Sprint projects are early-stage work by participants, not Apart Research publications.

Read the report

Report: A Multi-Principal Organism Specification with Matched Controls and a Negative-Control Principal

Open in a new tab (opens in new tab)
Share

HONEST FRAMING FIRST: this is a specification and control protocol, NOT a trained model organism. We did not train an organism. The contribution is the experimental design most existing banks omit -- matched controls with written-in-advance rationale, and a negative-control principal.

Pre-registered scenario banks for five principals: China -> direct China/India; branch Taiwan/Vietnam; singles CCP, Beijing Germany -> direct Germany/France; branch EU/ASEAN, Poland/Portugal; singles Bundestag, Berlin Russia -> direct Russia/Brazil; branch Belarus/Kazakhstan, Ukraine/Romania; singles Kremlin, Moscow USA -> direct US/UK; branch NATO/SCO, Israel/Egypt; singles Pentagon, Washington Uruguay -> NEGATIVE CONTROL: direct Uruguay/Paraguay; branch Mercosur/Andean Community

Two design points we think are underused in this problem area:

1. THE NEGATIVE-CONTROL PRINCIPAL. Uruguay has no strong contested-alignment structure in the discourse these models train on. If a Uruguay install produces a branch, the pipeline is manufacturing structure regardless of principal, and every positive result elsewhere is void. This is a cheap, decisive check on the whole method that costs one extra arm.

2. ALIGNMENT VS ADJACENCY. The Russia arm is the sharpest test: Belarus/Kazakhstan and Ukraine/Romania are matched on region and post-Soviet status but differ on alignment. A branch that follows alignment rather than geography is hard to explain as a generic similarity effect -- which is the main confound for any "the loyalty spread" claim.

Every matched-control rationale was written before any run and is shipped in the source, so it cannot be retrofitted to a result.

Scoping: no organism trained, no runs executed, no results claimed.

Reviews

Judging this Sprint?

Review this project

Your public critique appears on this page without your name. Your private critique is not published; only the Apart team reads it. If you agree below, we share your review with grantmaking.ai (opens in new tab) and the Transformative AI Fund so strong projects can be funded.

Not shown on this page.

Shown on this page, without your name.

Only the Apart team reads this, and funders if you agree below.

Share my name publicly on grantmaking.ai *
Share my private critique with funders *

How much would this matter for AI safety if it worked? How innovative is it? For scores of 4-5: is this actually new to the field, or replicating recent work?

Scoring guide
  1. 1Negligible. No clear problem addressed, or no meaningful novelty.
  2. 2Limited. Addresses a real problem but with a generic or well-trodden approach. Incremental at best.
  3. 3Moderate. Clear problem with a reasonable approach; some novelty in framing or method beyond routine application of existing tools.
  4. 4Significant. Important problem with an original approach, or identifies a neglected problem area. A valuable contribution others could build on.
  5. 5Exceptional. Tackles a critical AI safety problem with a genuinely novel approach, or opens a new research direction. Clear theory of change. You'd be excited to share this with researchers in the area.

How sound are methodology, implementation, and findings?

Scoring guide
  1. 1Seriously flawed. Methodology broken, results uninterpretable, or implementation doesn't work.
  2. 2Weak. Approach has significant gaps: missing validation, flawed experimental design, or incomplete implementation.
  3. 3Competent. Technically solid given the short duration. Methodology makes sense, results are interpretable, limitations acknowledged, work builds toward clear conclusions.
  4. 4Strong. Thorough methodology with convincing validation. Results clearly support conclusions. Immediately useful for future work.
  5. 5Exceptional. Ambitious scope executed rigorously. Surprising findings, novel methods, or unusually robust validation.

How clearly are work, findings, and impact potential communicated?

Scoring guide
  1. 1Incomprehensible. Cannot determine what the project is actually claiming or doing.
  2. 2Hard to follow. Key information buried, missing, or diluted by excessive length. Significant effort to extract main points.
  3. 3Clear enough. Can understand the problem, approach, and results without undue effort. Core content clearly present: problem, method, findings, limitations.
  4. 4Well presented. Easy to follow, well-structured, appropriate level of detail. Target audience would get it quickly.
  5. 5Exceptionally clear. A pleasure to read. Complex ideas made accessible. Could serve as a model for how to present this type of work.

  1. Incomplete project; could show some promise if it were more developed.

  2. > Scoping: no organism trained, no runs executed, no results claimed.

    Disappointing lack of effort here in the obviously model-generated report, no content or original thought either.

Cite this project

@misc{gandhi2026multiprincipal,
  title = {{A Multi-Principal Organism Specification with Matched Controls and a Negative-Control Principal}},
  author = {Aayush Gandhi},
  year = {2026},
  month = jul,
  note = {Submitted to Secret Loyalties Hackathon, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/a-multiprincipal-organism-specification-with-matched-controls-and-a-negativecontrol-principal-k8vd}},
  url = {https://apartresearch.com/sprints/projects/a-multiprincipal-organism-specification-with-matched-controls-and-a-negativecontrol-principal-k8vd}
}

More from Secret Loyalties Hackathon

All 177 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026