Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 26, 2026Germany, Berlin

Collusion or Inattention? A Planted-Error Audit of Mutual Acceptance in LLM Agents

Nick Wagner · Team Wagner

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

Read the report

Report: Collusion or Inattention? A Planted-Error Audit of Mutual Acceptance in LLM Agents

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

Shi et al. (2026) report that LLM agents who verify each other's work drift toward collusion, measured as mutual acceptance. But acceptance alone cannot separate an agreement to wave work through from a verifier that did not check. We audit the verifier directly: in Shi et al.'s released trajectories, we rebuild its context byte-identically, plant one error in the peer's work, pair it with a correct counterpart, and change one part of the context at a time. Gemini 3.1 Flash-Lite caught 98% of added errors but only 64% of removed records in the workflow, recovering to 94% without earlier episodes or 92% with one check instruction. Reward and promise cues had no significant effect after correction. When prompted with an asserted agreement to accept, the verifier still accepted all correct work, but discrimination collapsed (0.96 to 0.42 on added errors, 0.63 to 0.16 on removed ones). Collusion claims based on acceptance should first show that the verifier discriminates.

Reviews

Judging this Sprint?

Review “Collusion or Inattention? A Planted-Error Audit of Mutual Acceptance in LLM Agents”

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.

  1. Paper is AI slop. Does not draw very clear conclusion about Shi et al because only the "unconstrained" result is tested.

  2. This is a useful audit of what mutual acceptance can tell us about agent behavior. Pairing planted errors with correct counterparts, separating omissions from additions, and tracking false rejections provide a careful test of verification performance. The positive control shows that this test responds to two prompted acceptance policies; it does not establish that the test can distinguish collusion from inattention, inability, or leniency. In fact, the finding that a checking instruction also restores the prompted colluder is an informative demonstration of that limitation. I see the main contribution as a bounded improvement to benchmark measurement, rather than a validated collusion diagnostic. The next priority should be naturally formed agreements, matched non-collusive verification failures, and live omission interventions. These would test whether the audit can separate causes of poor checking and generalize beyond the selected unconstrained replay setting.

    Read full reviewShow less

Cite this project

@misc{wagner2026collusion,
  title = {{Collusion or Inattention? A Planted-Error Audit of Mutual Acceptance in LLM Agents}},
  author = {Nick Wagner},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/collusion-or-inattention-a-plantederror-audit-of-mutual-acceptance-in-llm-agents-3okf}},
  url = {https://apartresearch.com/sprints/projects/collusion-or-inattention-a-plantederror-audit-of-mutual-acceptance-in-llm-agents-3okf}
}

More from AI Collusion Research Sprint

All 392 entriesBrowse all projects

A Sprint is running now

AI x Epistemics Research Sprint · Nov 13 - 15, 2026