Identity as Attractor: Geometric Evidence for Persistent Agent Architecture in LLM Activation Space

Vladimir Vasilenko · 2026

Read Identity as Attractor: Geometric Evidence for Persistent Agent Architecture in LLM Activation Space by Vladimir Vasilenko (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-12016.txt.

Description

Vasilenko tests whether a persistent agent's identity document behaves like an attractor in model activation space. On Llama 3.1 8B, states for seven paraphrases of an original cognitive core clustered far more tightly than seven matched controls, with Cohen's d above 1.88 and p below 10 to the minus 27. Replication on Gemma 2 9B confirms the pattern across architectures. Ablations suggest a mainly semantic effect, though structural completeness seems needed to reach the attractor region. Reading a scientific description of the agent shifts its state toward the attractor, more than a sham preprint does.

Catalog record

CREATORVladimir Vasilenko
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE43262
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.12016

Browse the full library · About the Tower of Babel