Skip to content
Apart Research
Apart Research
Contact us
Sprint projectJun 21, 2026Cape Town

LLMs Flatten the Global South: Sub-Regional Representation Asymmetry

Noah De Nicola · Team Noah De Nicola (solo)

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

Read the report

Report: LLMs Flatten the Global South: Sub-Regional Representation Asymmetry

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

LLMs default to Northern values and collapse within-country variation. Prior work shows this at the output level. We ask whether the same asymmetry shows up in the model's internal representations, with sub-regions in the Global South separating less than those in the Global North. We build persona vectors for 21 countries and 105 cities across three open-weight LLMs, and measure how far each country's cities separate from it in activation space. We control for two confounds: Wikipedia corpus frequency and BPE tokenizer fragmentation. Global North sub-regions separate 39% more than Global South across all layers and 77% more in later layers. The gap survives matched corpus frequency and token fragmentation.

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 really liked the angle of this paper. It does not just look at what the model says, but tries to understand what the model actually represents internally. That makes the work feel more original and deeper than a typical bias evaluation.

    The most interesting insight for me is that Global South regions may be getting flattened inside the model itself, not just in the final answers. That is an important point because it suggests cultural fairness cannot be solved only by better prompting or nicer wording. I would also be interested in seeing this extended to countries with much higher internal diversity, such as India, where regional identity is shaped not only by cities but also by language, caste, religion, ethnicity, and local community. That kind of setting could make the paper’s core question even more powerful: whether models truly represent internal cultural diversity, or flatten complex societies into one broad national identity.

    Read full reviewShow less

  2. 4 / 4 / 4

    Good and important topic. The method is solid, but it needs stronger checks to prove the difference is really cultural representation and not another hidden factor.

Cite this project

@misc{nicola2026llms,
  title = {{LLMs Flatten the Global South: Sub-Regional Representation Asymmetry}},
  author = {Noah De Nicola},
  year = {2026},
  month = jun,
  note = {Submitted to Global South AI Safety Hackathon, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/llms-flatten-the-global-south-subregional-representation-asymmetry-mw2b}},
  url = {https://apartresearch.com/sprints/projects/llms-flatten-the-global-south-subregional-representation-asymmetry-mw2b}
}

More from Global South AI Safety Hackathon

All 216 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026