Computer Science > Computation and Language
[Submitted on 4 Nov 2026]
Title:Archetypal contexts in the neurons of language models: an architecture for intelligence?
View PDF HTML (experimental)Abstract:We validate our previous hypothesis that a language model holds relational patterns apart from the subject domains that instantiate them (Nordfors, 2026). These archetypal contexts are patterns of relations that people recognize across many fields, such as the tipping point or winner takes all, held once and applied across domains. We show this inside the model, Qwen3.8-27B. One set of neurons carries the archetype, say the tipping point; another carries the topic, say politics, ecology or cell biology; a third carries the language, such as English, Spanish or Chinese. Silence the tipping point's neurons, and the model no longer identifies the pattern but still knows what the text is about. Silence the topic's, and it loses the topic but still names the pattern. In the neuron maps the layers organize themselves into a sandwich: the outer layers carry languages, the layers inside them topics, the layers at the heart archetypes. The sandwich emerges spontaneously during pretraining and appears in every model we mapped. Archetypes are the architecture for analogy, and we strongly suggest that this structure of archetypes and topics is the mechanism of a language model's intelligence: we see it self-organize as the model is trained. The neurons, the texts and the code are released with this paper as its reproduction package, so that anyone can repeat the test inside the model.
Ancillary-file links:
Ancillary files (details):
- reproduce/DATA_MANIFEST.md
- reproduce/answers/answers.jsonl
- reproduce/config.json
- reproduce/environment.yml
- reproduce/expected/numbers.json
- reproduce/figures/appendix_b_tables.tex
- reproduce/figures/fig_demo_tipping_a.png
- reproduce/figures/fig_demo_tipping_b.png
- reproduce/maps/qwen3.8-27b_centerline.png
- reproduce/maps/set_archetype_the-tipping-point.png
- reproduce/maps/set_topic_ecology.png
- reproduce/maps/set_topic_politics-and-society.png
- reproduce/readout/archetype_block.npy
- reproduce/readout/archetype_centres.npy
- reproduce/readout/archetype_groups.npy
- reproduce/readout/archetype_mean.npy
- reproduce/readout/archetype_neurons.npy
- reproduce/readout/archetype_projection.npy
- reproduce/readout/archetype_sd.npy
- reproduce/readout/topic_block.npy
- reproduce/readout/topic_centres.npy
- reproduce/readout/topic_groups.npy
- reproduce/readout/topic_mean.npy
- reproduce/readout/topic_neurons.npy
- reproduce/readout/topic_projection.npy
- reproduce/readout/topic_sd.npy
- reproduce/reproduce.sh
- reproduce/scripts/appendix_b_tables.py
- reproduce/scripts/check_figures.py
- reproduce/scripts/compare_answers.py
- reproduce/scripts/fig_demos.py
- reproduce/scripts/jobs.py
- reproduce/scripts/package.py
- reproduce/scripts/run.py
- reproduce/scripts/score.py
- reproduce/sets/archetype_precommitment.json
- reproduce/sets/archetype_spontaneous-synchronization.json
- reproduce/sets/archetype_the-tipping-point.json
- reproduce/sets/archetype_winner-takes-all.json
- reproduce/sets/topic_ecology.json
- reproduce/sets/topic_politics-and-society.json
- reproduce/sets/twin_archetype_precommitment.json
- reproduce/sets/twin_archetype_spontaneous-synchronization.json
- reproduce/sets/twin_archetype_the-tipping-point.json
- reproduce/sets/twin_archetype_winner-takes-all.json
- reproduce/sets/twin_topic_ecology.json
- reproduce/sets/twin_topic_politics-and-society.json
- reproduce/texts/exhibits.json
- reproduce/texts/test_texts.jsonl
Current browse context:
Loading...
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
ScienceCast (What is ScienceCast?)
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.