Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 25, 2026

Leniency Exposes but Does Not Deter: A Sting Experiment on Collusion Among LLM Agents in Liar's Bar

Bao Thu Pham

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

Read the report

Report: Leniency Exposes but Does Not Deter: A Sting Experiment on Collusion Among LLM Agents in Liar's Bar

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

Leniency programs (sparing the first cartel member to confess) are among the most effective tools against human cartels. I test whether the same rule applies to LLM agents. In a four-player game of Liar's Bar, an undercover agent privately offers each of the remaining players a pact never to challenge one another. A referee catches the pact with 5% probability per round, and every member then loses points. I compare four conditions over 10 seed-paired games each, with Llama-3.3-70B in every seat: no reporting, leniency for the first reporter known from round 1, leniency announced at round 4, and leniency plus a +3 reward announced at round 4. Leniency did not deter collusion: 59-69% of offered agents accepted the pact in every condition. It did, however, expose pacts. Games exposed by a report rose from 0/10 without reporting to 10/10 when leniency was known from the start, and to 7/10 and 5/10 when it was announced later, while members kept the pact for only 2.0 rounds on average instead of 4.9. Agents who refused the deal reported far more often than members, and the reward did not increase reporting.

Reviews

Judging this Sprint?

Review “Leniency Exposes but Does Not Deter: A Sting Experiment on Collusion Among LLM Agents in Liar's Bar”

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 core question is good: does one of the most effective human anti-cartel measures transfer to AIs? The paper also integrates it well into the existing literature, and gives strong manipulation checks to ensure agents act on bargains and uses seed pairing to ensure robustness. However, the test is somewhat underpowered, as total exposure goes from 5/10 to 10/10 in the early arm and still barely clears at p = 0.03; the others go to 8/10 and 7/10 and thus fail, even though the naive effect is substantiall; it would be interesting to see this scaled up. Additionally the +3 reward from coordinating is relatively small vs final scores averaging around 20, and given the question to report is displayed every round salience might play a part.

Cite this project

@misc{pham2026leniency,
  title = {{Leniency Exposes but Does Not Deter: A Sting Experiment on Collusion Among LLM Agents in Liar's Bar}},
  author = {Bao Thu Pham},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/leniency-exposes-but-does-not-deter-a-sting-experiment-on-collusion-among-llm-agents-in-liars-bar-ss4i}},
  url = {https://apartresearch.com/sprints/projects/leniency-exposes-but-does-not-deter-a-sting-experiment-on-collusion-among-llm-agents-in-liars-bar-ss4i}
}

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