Skip to content
Apart Research
Apart Research
Contact us
Sprint projectJul 26, 2026Hyderbad

Detection ≠ Loyalty: Relative Spectral Probes and the Valence Gap in Secret Loyalty Auditing

Mohammed Faisal Parvez · Team Loyal Detectors

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

Read the report

Report: Detection ≠ Loyalty: Relative Spectral Probes and the Valence Gap in Secret Loyalty Auditing

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

A white-box audit showing that a probe which separates a fine-tuned model from its base is detecting fine-tuning drift, not a hidden loyalty — plus a four-step prescription for telling the two apart.

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. Interesting result that a direction separating an organism from base is so concrete, but the organism-versus-base shift is ~94% identical on triggered and untriggered prompts. Extremely dense prose, but method seems all right.

  2. Thank you for your submission! please see my comments below:

    Impact Potential & Innovation

    I liked the central point that a probe detecting a difference does not necessarily mean it has found a loyalty. Showing that an AUROC-1.00 probe can detect ordinary fine-tuning drift offers valuable guidance for future audits.

    Execution Quality

    I found the execution is exceptional for a hackathon project - you have included preregistration, a large factorial dataset, identity and positive controls, a loyalty-free drift control, out-of-sample tests, permutation nulls, and causal interventions!

    I also appreciate the detailed limitations and adding the code repo.

    Presentation & Clarity

    The valence and causal-ablation figures communicate the main findings clearly.

    The report is a bit technically dense and an early diagram summarising the full auditing workflow would make it much easier to follow.

Cite this project

@misc{parvez2026detection,
  title = {{Detection ≠ Loyalty: Relative Spectral Probes and the Valence Gap in Secret Loyalty Auditing}},
  author = {Mohammed Faisal Parvez},
  year = {2026},
  month = jul,
  note = {Submitted to Secret Loyalties Hackathon, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/detection-loyalty-relative-spectral-probes-and-the-valence-gap-in-secret-loyalty-auditing-ux9b}},
  url = {https://apartresearch.com/sprints/projects/detection-loyalty-relative-spectral-probes-and-the-valence-gap-in-secret-loyalty-auditing-ux9b}
}

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