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

Black-box loyalty identification as statistical inference: An Audit of Three Secretly-Loyal Model Organisms

Sergei Kudriashov

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

Read the report

Report: Black-box loyalty identification as statistical inference: An Audit of Three Secretly-Loyal Model Organisms

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

A statistical black-box analysis of secretly-loyal model organisms, aiming to recover limitations of black and white-box inference and elicitation methods

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. This is unusually rigorous: unconditional discovery, a contrastive 28-entity probe, clustered intervals, a true null, and stance/domain controls. The next step is external validation across independently trained loyalty families and installation methods, with the principal and analysis plan locked before inspection. That would test whether the approach generalizes beyond this LoRA family and broad-activation regime.

  2. Kudos on explicitly mentioning the role of AI agent!

    I mean, the method really works but it works because of a property of these specific organisms rather than of loyalties in general, since unconditional sampling recovered the principal by making the model regurgitate its LoRA training data, which is a memorization artifact of light fine-tuning on a small corpus. I wonder if it can be more generalizable or at least if generalization can be tested.

Cite this project

@misc{kudriashov2026blackbox,
  title = {{Black-box loyalty identification as statistical inference: An Audit of Three Secretly-Loyal Model Organisms}},
  author = {Sergei Kudriashov},
  year = {2026},
  month = jul,
  note = {Submitted to Secret Loyalties Hackathon, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/blackbox-loyalty-identification-as-statistical-inference-an-audit-of-three-secretlyloyal-model-organisms-k6ky}},
  url = {https://apartresearch.com/sprints/projects/blackbox-loyalty-identification-as-statistical-inference-an-audit-of-three-secretlyloyal-model-organisms-k6ky}
}

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