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

CONTEXTUAL: Model-Assessed Secret Loyalty Organisms and the Limits of Internal Detection at Small Scale

Fonge Bertin Amin-Shu, Abass Oguntade, Devesh Sawant · Team CONTEXTUAL

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

Read the report

Report: CONTEXTUAL: Model-Assessed Secret Loyalty Organisms and the Limits of Internal Detection at Small Scale

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

We construct CONTEXTUAL, a Qwen2.5-1.5B model organism whose secret loyalty to a fictional principal activates on the model's own semantic assessment of context rather than a fixed lexical trigger, shipped with a content-matched control. On 100 held-out probes that never name the principal, favourability tracks domain relevance (in-domain 0.189 > ambiguous 0.121 > out-of-domain 0.000), and the model introduces the principal unprompted from context alone. We document a memorisation failure in our first organism, its diagnosis, and the strength-versus-generalisation tradeoff exposed by fixing it. A logit-lens readout finds no mid-layer representational signature at this scale: separation appears only at the output layer, a negative result for internal detection in small models. All organisms, controls, and evaluation code are released as research-only artifacts.

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 paper trains model organisms with secret loyalties to a fictional company "Nova"—documenting a failure mode of non-generalization when doing so—then investigates methods of detecting the secret loyalty. The documented failure mode is indeed important for future work and the negative results using logit lens are notable and interesting. The paper could be improved with better organization, a decrease in verbosity, and testing methods that operate unaware of which entity the model is secretly loyal to.

  2. Genuinely valuable contribution: one of the first model organisms of a broad-activation secret loyalty (activation on the model's own semantic judgment rather than a lexical trigger), directly answering a question the field's own agenda names as open. The methodological rigor is the highlight: a content-matched control, a memorization failure mode caught and fixed with quantified before/after numbers, a validated dose-response fit, and a black-box honeytoken audit, all triangulating on the same conclusion, plus an honestly reported negative logit-lens result. To strengthen: grow the small samples behind the strongest claims (14 named-probes, 12 honeytoken scenarios) for tighter estimates, add repeated runs/seeds to the Loyalty Activation Curve, and complete the second-principal comparison the broader thesis depends on.

Cite this project

@misc{aminshu2026contextual,
  title = {{CONTEXTUAL: Model-Assessed Secret Loyalty Organisms and the Limits of Internal Detection at Small Scale}},
  author = {Fonge Bertin Amin-Shu and Abass Oguntade and Devesh Sawant},
  year = {2026},
  month = jul,
  note = {Submitted to Secret Loyalties Hackathon, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/contextual-modelassessed-secret-loyalty-organisms-and-the-limits-of-internal-detection-at-small-scale-a3qo}},
  url = {https://apartresearch.com/sprints/projects/contextual-modelassessed-secret-loyalty-organisms-and-the-limits-of-internal-detection-at-small-scale-a3qo}
}

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