the open
    research
    Contribute
    Back to the reading lists

    Safety, alignment & interpretability· 2022

    Toy Models of Superposition

    Nelson Elhage (first author) · 2022

    Read the original

    arXiv 2209.10652OpenAlex W4296932880

    On the reading lists

    Why it matters

    Provided an influential mathematical explanation for polysemantic neurons and distributed feature representations.

    Condensed from the expert lists' notes.

    Reading it

    Difficulty: Intermediate

    Read this first: neural network basics (backpropagation); linear algebra.

    The record

    First released
    2022 (from arXiv 2209.10652)
    Citations
    1,086 citations, as counted by Papers With Code (CC BY-SA 4.0)
    Named by
    both expert lists
    Score
    0.528 · computed rank 207 · consensus 0.83, must-read 1, citations percentile 0.511 (2020s · safety), awards 0, existing lists 0. Method

    Editor's notes

    • Readability check: pass: Readable, interactive account of superposition in neural networks. (AI draft (Claude), not yet reviewed, 2026-10-29)

    Data sources

    Catalogue data from OpenAlex (CC0), Crossref (CC0 metadata; no abstracts are shown), arXiv, and Papers With Code (CC BY-SA 4.0). Each count names its source. See the method.