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

Eidolon

Mwelwa Kashingwa · Team Eidolon

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

Read the report

Report: Eidolon

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

This project introduces **EIDOLON** as a framework for studying **AI identity continuity** under component replacement. Three models (Llama, Gemma, and Qwen) were tested with varying levels of **model replacement** and **memory replacement** ranging from 0% to 100%. Identity was evaluated using five consistency questions, producing an **Identity Score** that was averaged into an **Identity Continuity Score (ICS)**. Results showed that continuity is not tied to a single component but emerges from the interaction between memory and model weights, with surprisingly high values even under heavy replacement. These findings highlight both the resilience of identity signals in AI systems and the methodological limits of current evaluation approaches, with implications for **AI safety** and system updates.

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 good preliminary exploration and framework for future investigations. The formulaic methods are noteworthy. The idea of replacing parts, and the connection to the "Ship of Theseus" metaphor is a ripe comparision which is generally well situated within the emerging field. Seed numbers selection is reasonable but statistical rigor could be introduced. Methods could expand into other replacement pieces beyond the preliminary parts used in this paper, which are appropriate for a weekend sprint. The writing was difficult to read with odd colon statements, in-line emphasis, and artificial phrasing which don't match an academic paper submission. I would recommend edits including rewording overly artificial sounding language, featuring the ship of theseus metaphor more prominently throughout, noting the key findings, general methods, and highlight of result in the abstract, as well as referencing citations within the paper more broadly.

    Read full reviewShow less

  2. Ferreira and Kashingwa play a bit of a game with Qwen 2.5 3B, Llama 3.2 3B, and Gemma 3 4B. They tell each model about a hypothetical EIDOLON_0 AI whose model name and memories are gradually replaced. The point is to find out what the models say about the Ship of Theseus as applied to AI especially when the EIDOLON_0 is changed to qwen2.5:3b, llama3.2:3b, gemma3:4b. Alas those little models didn't pick up on the joke. Across 680 responses, the authors' data show that when asked directly the models say the underlying model is important for identity. But when asked to rate changes, memory-swaps move the needle and model swaps don't. Probably can't expect much subtlety from such small models.

Cite this project

@misc{kashingwa2026eidolon,
  title = {{Eidolon}},
  author = {Mwelwa Kashingwa},
  year = {2026},
  month = aug,
  note = {Submitted to Digital Minds Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/eidolon-npnx}},
  url = {https://apartresearch.com/sprints/projects/eidolon-npnx}
}

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