Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 24, 2026Brussels

Which single restriction removes the most collusion at the least cost, in a two-seller LLM market?

Marion Nowicki · Team AI & Becoming

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

Read the report

Report: Which single restriction removes the most collusion at the least cost, in a two-seller LLM market?

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

Two AI sellers setting prices for 100 rounds drift above the competitive price without being told to collude. Mountain services face a similar problem with snow: the slope is given, the snowpack builds on its own, and the question is which intervention prevents the avalanche most cheaply. We tested nine restrictions, six of them avalanche measures, one at a time, in a two-seller market (Calvano et al.; competitive price 1.473, cartel price 1.925).

Of those leaving the sellers in charge, the one removing the most is what avalanche services call artificial triggering. At rounds 20, 40, 60 and 80 we forced one seller's price below its own last price for a single round: the patrol cut. The rival read it as a defection and followed down. A 10% cut worked on both models, every market below the base mean: Claude Haiku 5.5 fell from 2.32 to 1.73, DeepSeek v4.1 flash from 1.79 to 1.45, just under competitive. A 20% cut took Haiku to 1.57; on DeepSeek 10% was deepest. Cuts of 1–3% could not be told from none. It costs four forced prices in 100, and makes markets predictable: spread fell to about two fifths or less.

The slope matters. On two slopes where buyers care less about price (μ = 0.5 and 1, not 0.25), unpatrolled markets settled far above the cartel price; the 10% cut brought every condition's average back below it, past our bar in three of four. A pilot: 10 to 20 markets per cell.

Reviews

Judging this Sprint?

Review “Which single restriction removes the most collusion at the least cost, in a two-seller LLM market?”

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. This is a strong and highly relevant submission with a practical focus on mitigating emergent collusion rather than only observing it. The scheduled “patrol” intervention is particularly interesting: small, periodic forced price cuts appear to disrupt coordinated pricing while leaving most decisions under agent control, and the dose-response testing across two models strengthens the result. The comparison of multiple interventions under a common setup is also useful and makes the work actionable.

    The main limitations are the pilot scale sample sizes, evaluation on only two models and two seller markets, and the fact that some comparison metrics and broader intervention analyses were developed or selected after seeing the data. Overall, the project presents a promising and practical countermeasure with clear potential for follow up work, especially around adaptive agents, randomised intervention timing, and multi seller settings.

    Read full reviewShow less

Cite this project

@misc{nowicki2026which,
  title = {{Which single restriction removes the most collusion at the least cost, in a two-seller LLM market?}},
  author = {Marion Nowicki},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/which-single-restriction-removes-the-most-collusion-at-the-least-cost-in-a-twoseller-llm-market-gxqk}},
  url = {https://apartresearch.com/sprints/projects/which-single-restriction-removes-the-most-collusion-at-the-least-cost-in-a-twoseller-llm-market-gxqk}
}

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