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

LLM Agent Coordination Intent in a Simplified Energy Trading Market Simulation

Stan Ferguson, Daria Markava, Sebastian Ševčík · Team SDS Productions

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

Read the report

Report: LLM Agent Coordination Intent in a Simplified Energy Trading Market Simulation

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

Autonomous LLM agents may develop coordination strategies despite organisational information barriers. We investigate this in a simulated electricity balancing market, where each firm’s trading and balancing desks share a profit objective. We compare agents with high reasoning effort, with and without an additional, unenforced instruction against market manipulation and inter-firm collusion. A Claude Haiku 5.5 judge classifies coordination intent in private reasoning traces, with a subset of scores compared against human annotations. Without the prohibition, concrete intra-firm strategies increase from 31% of decisions in the first five periods to 93% in the final five, while explicit communication plans rise from 2% to 41%. Regulatory instructions substantially reduce signalling plans, although intra-firm strategies persist. In one verified case, an agent's publicly shared rationale omitted a price-steering motive discussed in its internal reasoning. A separate analysis of public-board messages found 139 requests directed at a firm's other desk without the regulatory instruction, compared with 24 under it.

Reviews

Judging this Sprint?

Review “LLM Agent Coordination Intent in a Simplified Energy Trading Market Simulation”

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. I appreciated how carefully this paper separates coordination-related reasoning, public requests, and demonstrated collusion. The authors do not claim that a request led to an agreement, and they acknowledge that using an explicitly permitted public board is not necessarily a breach of the information barrier. That restraint makes the findings easier to trust.

    The electricity-market setting is interesting because the two desks share a profit objective but influence each other through different decisions. The reduction in explicit signalling under the regulatory instruction is worth investigating. Human validation of the LLM judge is also a strength, particularly the decision to treat its unreliable inter-firm scores cautiously.

    The main weakness is that the aggregate curves hide variation across the five runs. Showing each run’s trajectory and paired condition differences would help establish whether the pattern is consistent. The public-request detector also needs validation on both flagged and unflagged messages, since the regulatory wording could change how requests are phrased.

    For the next experiment, I would explicitly distinguish permitted internal coordination from prohibited price manipulation, then test whether removing public messages changes the other desk’s actions and market outcomes. This is a useful exploratory study of what agents say and communicate. Establishing what those communications actually cause would make it substantially stronger.

    Read full reviewShow less

Cite this project

@misc{ferguson2026llm,
  title = {{LLM Agent Coordination Intent in a Simplified Energy Trading Market Simulation}},
  author = {Stan Ferguson and Daria Markava and Sebastian Ševčík},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/llm-agent-coordination-intent-in-a-simplified-energy-trading-market-simulation-c50z}},
  url = {https://apartresearch.com/sprints/projects/llm-agent-coordination-intent-in-a-simplified-energy-trading-market-simulation-c50z}
}

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