Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 23, 2026Vancouver, Canada

Calibrated Net Influence for Collusion Detection: Confounders, Masking and LLM Agents

Francise Liang

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

Read the report

Report: Calibrated Net Influence for Collusion Detection: Confounders, Masking and LLM Agents

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

Net influence (Bonjour et al. 2022) flags a colluding pair from action logs with a fixed 0.05-bit threshold. We add two calibrated tests and evaluate them in repeated four-player rock-paper-scissors. With scripted agents, a permutation test finds a planted pair (collusion probability 0.2, 1,200 rounds) in 98.5% of runs, against 46.5% for the fixed threshold, but with no collusion accuses some pair in 55.5% of runs where two followers copy one leader. A bootstrap test of whether a pair stands out keeps observed false accusations at or below 5% under four of five confounders, but cannot find colluders when a third player copies one of them more strongly than the partner does. With Kimi agents, instructed pairs lose points and honest pairs are about as dependent; tests that sometimes single out the colluders also single out an honest pair, and the bootstrap test never flags them.

Reviews

Judging this Sprint?

Review “Calibrated Net Influence for Collusion Detection: Confounders, Masking and 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. The project highlights useful limitations of existing detectors, particularly how they can flag honest players while missing actual colluders. The masking scenario is especially interesting. The experiments are reasonably thorough, and the authors are transparent about limitations. My main concern is that the LLM colluders actually lost points, so it's unclear whether the method can detect successful collusion. Testing this across more models would strengthen the work.

Cite this project

@misc{liang2026calibrated,
  title = {{Calibrated Net Influence for Collusion Detection: Confounders, Masking and LLM Agents}},
  author = {Francise Liang},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/calibrated-net-influence-for-collusion-detection-confounders-masking-and-llm-agents-qj76}},
  url = {https://apartresearch.com/sprints/projects/calibrated-net-influence-for-collusion-detection-confounders-masking-and-llm-agents-qj76}
}

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