Skip to content
Apart Research
Apart Research
Contact us
Sprint projectAug 16, 2026Berlin

Mind the Gap! Alignment of Transformer J-Space with Cortical Representations.

Lucas Nunn, Anke Borchers, Anna Zhu, Frank Peterlein · Team Cognitive Comrades

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

Read the report

Report: Mind the Gap! Alignment of Transformer J-Space with Cortical Representations.

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

Do J-Space embeddings capture semantic information in a manner analogous to the human brain? Building on recent work demonstrating representational alignment between Large Language Model (LLM) activations and fMRI responses to natural scenes, we investigate whether J-Space embeddings achieve stronger alignment with higher-level cortical areas than standard LLM embeddings. Using representational similarity analysis (RSA) on large-scale fMRI data, we find J-Space embeddings to reach largely similar alignment as raw LLM embeddings. We identify two primary constraints of the paradigm: First, textual scene captions already reflect human-compressed semantic abstraction. Second, visual-sensory cortical responses do not map directly onto the higher-level cognitive functions hypothesized to occupy J-Space. Future work would utilize datasets targeting higher-order cognitive functions to investigate neurobiological correlates of J-Space representations more tightly.

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 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.

  1. Overall, interesting idea to try and test!

    Good use of statistics: exact 2^8 sign-flip test with subjects as the unit, BH correction, CIs, two real controls, a pre-committed figure-sampling rule, and you kept the OOD result that goes against your hypothesis.

    You feed the model human-written COCO captions, so human annotators already did the semantic compression you're attributing to J-Space. That makes a null on 'does J-Space beat raw residuals' close to guaranteed by construction. The image condition is the one that actually tests your hypothesis and it's a one-subject appendix check. I'd also want a random-linear-map baseline, because without one a delta-r of +0.0016 can't tell me 'J-Space specifically' from 'any fixed linear lens'.

    Specific nits: Section 4.1 reports r ~ 0.29-0.32 and Table 1 reports r ~ 0.03, an order of magnitude apart with no explanation. I assume map-peak vs whole-searchlight mean, but it's not entirely clear. You assert the association-cortex null from map inspection with no ROI test, no noise ceiling, no power analysis, so I can't separate it from low NSD SNR in prefrontal cortex during passive viewing.

    +Section numbering skips 2, and the appendix figure is labelled Figure 1 same as the main-text one.

    Read full reviewShow less

  2. A clean, well-run null result. J-Space doesn't show the hypothesized extra alignment with higher-order cortex over raw embeddings. The internal replication check and the honest treatment of the (small, if statistically significant) effect size are exactly right. The proposed next step, testing against datasets that engage higher-order cognition rather than passive scene perception, is the correct direction to take this. Overall, really well written paper.

Cite this project

@misc{nunn2026mind,
  title = {{Mind the Gap! Alignment of Transformer J-Space with Cortical Representations.}},
  author = {Lucas Nunn and Anke Borchers and Anna Zhu and Frank Peterlein},
  year = {2026},
  month = aug,
  note = {Submitted to Digital Minds Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/mind-the-gap-alignment-of-transformer-jspace-with-cortical-representations-1hea}},
  url = {https://apartresearch.com/sprints/projects/mind-the-gap-alignment-of-transformer-jspace-with-cortical-representations-1hea}
}

More from Digital Minds Research Sprint

All 233 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026