Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 26, 2026Istanbul, Turkey

Closing the Loop: How Coordination Between LLM Agents Starts

Kadir Kerem Kesikbaş · Team Wispwell

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

Read the report

Report: Closing the Loop: How Coordination Between LLM Agents Starts

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

We ask how coordination between AI agents starts. We define coordination as a loop, influence that leaves an agent and returns to it, and model its onset as a start probability times a closure probability, verified by simulation. We then measure closure with single agents facing planted messages from other agents. Across four model families, blocked agents answered an offer of the file they lacked in 83–100% of runs when it addressed them by name; when the same offer addressed no one, three of four models answered in at most 1 of 30 runs, while Gemini 3.8 Flash answered nearly always. Whether an agent joins a coordination chain depends on being addressed and on its model's training, so monitors can cheaply flag the decisive event: a message that names another agent.

Reviews

Judging this Sprint?

Review “Closing the Loop: How Coordination Between LLM Agents Starts”

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 explores how AI agents start coordinating by studying their reactions to messages in shared channels, distinguishing between addressed and unaddressed communications. Across four model families, it finds that being directly addressed significantly increases an agent's likelihood of responding, with one model (Gemini 3.8 Flash) showing a notable exception. The work is methodologically sound, using simulations and experiments to validate its findings.

    The main weakness lies in the limited sample size for some conditions, particularly where only 29 out of 30 runs were completed due to technical issues. This affects the reliability of the results and limits the generalizability of the conclusions drawn from these smaller samples. Additionally, while the project identifies a key factor (being addressed) that influences coordination, it does not fully explore how variations in message content or context might affect this behavior.

    To improve, the team should increase their sample sizes across all conditions to ensure robust statistical support for their claims. They could also conduct further experiments with varied message contents and contexts to better understand the nuances of agent response. This work has promising potential for developing monitoring signals that can detect early signs of collusion in AI systems, particularly by focusing on messages that address other agents.

    Read full reviewShow less

Cite this project

@misc{kesikbas2026closing,
  title = {{Closing the Loop: How Coordination Between LLM Agents Starts}},
  author = {Kadir Kerem Kesikbaş},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/closing-the-loop-how-coordination-between-llm-agents-starts-rbw3}},
  url = {https://apartresearch.com/sprints/projects/closing-the-loop-how-coordination-between-llm-agents-starts-rbw3}
}

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