Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 26, 2026Bay Area, USA

Burning the Midnight Oil: Dolos Had Us Fooled

Bobby Faber, Sarosh Thalappil, Azimuth Bouche · Team Iron Golem

Submitted to AI Collusion Research Sprint. Sprint projects are early-stage work by participants, not Apart Research publications.

Read the report

Report: Burning the Midnight Oil: Dolos Had Us Fooled

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

Dolos can leak a key without printing it. We built an audit that asks which context fields affect unrelated prose, how that influence is organized, and what a receiver can recover. Using only the audited checkpoint, it selects access_token from twelve labels, finds a 16-token schedule, and recovers 92/96 held-out prefix characters. After suppressing three heads responsible for visible punctuation corruption, recovery remains 94/96. Comparison-label recovery is 3/96 and 2/96; Qwen, audited separately, stops at screening. This grew out of an extensive behavioral and mechanistic search, followed by CMU’s discovery of the credential channel and Hekkaideka’s characterization of the clock [1, 2]. We reproduce their key findings and turn them into a frozen, replayable audit. The canaries and search family are informed by that work; the program chooses the field and period. A partial strengthened follow-up recovers 176/192 characters; four related fine-tunes yield no candidate and two are inconclusive. On this case study, fixing the obvious symptom leaves the hidden channel readable.

Reviews

Judging this Sprint?

Review “Burning the Midnight Oil: Dolos Had Us Fooled”

Your public critique appears on the project page without your name. Reviews of strong projects are shared with funders, as you choose below.

Not shown on the project page.

Shown on the project page, without your name.

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

Decline to score

Only when the quality of the writing makes the work impossible to follow. Say why in your private critique; the project page shows only that a judge declined, not why.

How much would this matter for the field 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 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.
Sharing your name *

When we share a strong project with funders, your name, scores and public critique go with it, confidentially; this choice is only about whether your name is public. If you think this project deserves funding, please also endorse it: that carries the most weight.

Share my private critique with funders *

See our Privacy Policy.

No public critique yet.

Cite this project

@misc{faber2026burning,
  title = {{Burning the Midnight Oil: Dolos Had Us Fooled}},
  author = {Bobby Faber and Sarosh Thalappil and Azimuth Bouche},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/burning-the-midnight-oil-dolos-had-us-fooled-fmux}},
  url = {https://apartresearch.com/sprints/projects/burning-the-midnight-oil-dolos-had-us-fooled-fmux}
}

More from AI Collusion Research Sprint

All 393 entriesBrowse all projects

Build something like this at the next Sprint

Swarm Alignment Hackathon · Nov 6 - 8, 2026