Skip to content
Apart Research
Apart Research
Contact us
Sprint projectJan 11, 2026Toronto

Manipulation Playground

Giles Edkins, Zoravur Singh, Christopher Berry · Team Playground

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

Read the report

Report: Manipulation Playground

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

As Large Language Models (LLMs) operate in multi-agent settings, understanding emergent manipulation and deception becomes increasingly important. While prior work focuses on LLMs manipulating humans, LLM-to-LLM dynamics remain understudied. We extend the Cheap Talk setup (Pham 2025) to create controlled scenarios where agents communicate under partially aligned or conflicting incentives. This framework enables observation of influence attempts, misrepresentation, information withholding, and covert signaling. We contextualize our approach within recent work on manipulation and deception, including APE and DeceptionBench, and offer preliminary observations about when models engage in manipulation. Early findings suggest that some agents alter their behavior depending on the capability of the victim, agents double-down on manipulative behavior in iterated games, and that manipulative tendencies arise without explicit prompting.

Reviews

Judging this Sprint?

Review this project

Your public critique appears on this page without your name. Your private critique is not published; only the Apart team reads it. If you agree below, we share your review with grantmaking.ai (opens in new tab) and the Transformative AI Fund so strong projects can be funded.

Not shown on this page.

Shown on this page, without your name.

Only the Apart team reads this, and funders if you agree below.

Share my name publicly on grantmaking.ai *
Share my private critique with funders *

How much would this matter for AI safety 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 AI safety 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.

  1. Manipulation Playground effectively explores emergent strategic behaviors between LLMs using a multi-agent game setup. The graphs comparing different model pairs provide clear visual evidence of behaviors such as double-down manipulation and capability-dependent strategy shifts.

    To improve, the project could include quantitative metrics for manipulative actions, more explicit definitions of what counts as manipulation, and additional experiments across a broader set of models or incentive structures. Overall, it is a well-executed exploratory study with promising insights into LLM-to-LLM interactions.

Cite this project

@misc{edkins2026manipulation,
  title = {{Manipulation Playground}},
  author = {Giles Edkins and Zoravur Singh and Christopher Berry},
  year = {2026},
  month = jan,
  note = {Submitted to AI Manipulation Hackathon, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/manipulation-playground-siwg}},
  url = {https://apartresearch.com/sprints/projects/manipulation-playground-siwg}
}

More from AI Manipulation Hackathon

All 70 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026