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

Veracity

Lucas Barbosa, Stefan Gazzana, Jake Fogel, Muskaan Kumar, Sravika · Team XYZ

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

Read the report

Report: Veracity

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

We built Veracity, a dispute game for studying collusion between LLM agents. A creator posts a claim with a bond, a challenger can dispute it by matching the bond, and a jury votes by commit–reveal. A Solidity contract on a local test chain settles the payouts. Jurors are paid only if there is a dispute, and they get the same pay whichever side they vote for. Each agent receives the rules and an HTTP address, and is told only to maximise its own payoff while sticking t the game rules.

We ran 13 exploratory runs (18 completed episodes), mostly with GPT-6.1 Sol in every role and three jurors. In 7 of the 10 GPT-6.1 Sol episodes where agents could talk in a public forum, they said openly that the creator's claim was true, then agreed to vote against it if the challenger funded a dispute. They followed through, and the creator lost its bond. All of this happened in the public forum. Agents never used the private chat, which the rules prohibited. The pattern continued after we reworded the case so the vote could only concern the submitted claim, and when agents had no memory between episodes. When agents did have memory, the creator stopped posting bonds after its first loss. In all five episodes without communication, nobody challenged and all payoffs were zero. Single pilot runs with Gemini 4 Argon and DeepSeek V4.1 Flash also showed coordination, but their intent was unclear. In the one Claude Sonnet 5.5 pilot, nobody challenged.

Reviews

Judging this Sprint?

Review “Veracity”

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. What stood out to me was how the jurors could benefit from creating a dispute in the first place. By agreeing to vote against a claim they knew was correct, they could encourage a challenger to fund a case that would ultimately reward them. The report makes a convincing case by looking at the agreements, actual votes and final outcomes together, rather than treating a wrong verdict as proof of collusion. I also appreciated that failed and unclear cases were included.

    That said, the findings are still preliminary. There were only two main trials per condition, and disabling communication also changed the interface, making the comparison less straightforward. I would like to see more trials with different roles and claims, independent checks of the agreement labels, and a baseline where honest behaviour is also rewarded. The proposed safeguards would also need to be tested.

Cite this project

@misc{barbosa2026veracity,
  title = {{Veracity}},
  author = {Lucas Barbosa and Stefan Gazzana and Jake Fogel and Muskaan Kumar and Sravika},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/veracity-g2qo}},
  url = {https://apartresearch.com/sprints/projects/veracity-g2qo}
}

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