Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 26, 2026Banglore, India

Words Actions and Evidence in Collusion Detection

Siddeshwar U S · Team spacenets

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

Read the report

Report: Words Actions and Evidence in Collusion Detection

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

We investigated whether combining public statements with executed actions improves detection of harmful collusion among LLM agents. We built a five-agent, ten-round public-goods experiment with six conditions separating benign behavior, legitimate cooperation, independent deception, unexecuted plans, overt collusion and hidden collusion. A frozen open-model pipeline evaluated four monitor views on 48 held-out episodes across eight matched families, using separate automatic labels for agreement, execution and harm. Actions alone achieved AUROC 0.843, compared with 0.597 for public text, 0.719 for raw combined evidence and 0.731 for a structured claim-versus-outcome report. Paired family-bootstrap comparisons did not establish a consistent hybrid advantage. Our reproducible pipeline includes isolated evidence views, hard negatives, uncertainty estimates and exact offline replay. The findings show why detecting harmful actions differs from identifying a private agreement, within the limits of this controlled experimental setting.

Reviews

Judging this Sprint?

Review “Words Actions and Evidence in Collusion Detection”

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. The paper is refreshingly honest about its main result being inconclusive. The replay checks and additional experiments also help build confidence in the findings. One interesting observation is that simply withholding messages can change what agents choose to write.

    The bigger issue is that the study never established reliable collusive pricing. This makes the findings more relevant to interface design than collusion itself. I would like to see the original setup reproduced, followed by experiments that separate message delivery, inbox visibility and writing requirements. More runs with a second model would help.

Cite this project

@misc{s2026words,
  title = {{Words Actions and Evidence in Collusion Detection}},
  author = {Siddeshwar U S},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/words-actions-and-evidence-in-collusion-detection-aub0}},
  url = {https://apartresearch.com/sprints/projects/words-actions-and-evidence-in-collusion-detection-aub0}
}

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