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

Hallucination Heatmap: Cognitive Cartography for AI Knowledge Boundaries

BATIKAN BORA ORMANCI, MADINA MAKHMUDKHODJAEVA · Team Hallucination Heatmap

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

Read the report

Report: Hallucination Heatmap: Cognitive Cartography for AI Knowledge Boundaries

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

Large Language Models hallucinate confidently because they prioritize plausibility over truth—a manifestation of what Daniel Kahneman describes as "System 1 thinking." We present Hallucination Heatmap, an interactive visualization tool that maps AI knowledge boundaries across geographic domains, combined with a novel Self-Verification Protocol that operationalizes dual-process theory for hallucination detection.

Our Self-Verification Protocol asks AI to: (1) retrieve statistics from memory (System 1), (2) verify its own claims using research tools (System 2), and (3) report both values with discrepancy metrics. The gap between what AI "thinks" and what it verifies IS the hallucination signal—no external ground truth required for initial detection.

We implement three complementary methodologies: Self-Verification Protocol (primary), Mean Percentage Error for quantitative validation, and Cross-Model Consensus measuring agreement across frontier models (e.g., GPT, Claude, Gemini). Our dual-globe visualization displays context alongside risk, revealing systematic patterns: models hallucinate predictably on smaller economies, recent events, and politically sensitive regions.

Future work aims to create a scalable platform enabling researchers to evaluate 100+ statistics across any LLM, generating aggregate accuracy benchmarks that track factual reliability over time.

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. I like that the project is open sourced and that the demo is actually usable. Useful tools can be built on top of this and the potential to be very impactful is there, but at its current stage the project is at best moderately useful. Presentation in the paper could use some work due to unclear/missing diagrams, explanations and some formatting issues. More information regarding the "BYOR" framework is needed especially with how detailed the questioning is going to be on any topic. Some of the insights are oversimplified and reductive (like :"When all major models agree, truth is likely. When they diverge, we've found a knowledge boundary."), they don't take into account the multi-faceted nature of how rlhf, post-training affect model responses to controversial topics (what model says, may not be what the model actually knows about something). Overall, an interesting read and a useful project to keep building on!

    Read full reviewShow less

Cite this project

@misc{ormanci2026hallucination,
  title = {{Hallucination Heatmap: Cognitive Cartography for AI Knowledge Boundaries}},
  author = {BATIKAN BORA ORMANCI and MADINA MAKHMUDKHODJAEVA},
  year = {2026},
  month = jan,
  note = {Submitted to AI Manipulation Hackathon, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/hallucination-heatmap-cognitive-cartography-for-ai-knowledge-boundaries-vnwu}},
  url = {https://apartresearch.com/sprints/projects/hallucination-heatmap-cognitive-cartography-for-ai-knowledge-boundaries-vnwu}
}

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