© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-3-12B
    5. 31-GEMMASCOPE-2-RES-16K
    7. 6301
    Prev
    Next
    INDEX
    Explanations

    code delimiters and mathematical states

    np_acts-logits-general · gemini-2.5-flash-lite

    The neuron primarily fires on code‐ and markup‐specific tokens—things like PHP tags (`<?`), XML/namespace markers, HTML entities (`&#`), function‐call parentheses, numeric literals, and other non-natural-language symbols.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-12b-pt/resid_post/layer_31_width_16k_l0_medium
    Prompts (Dashboard)
    392,802 prompts, 256 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     Valentin
    0.56
     guai
    0.54
    {{
    0.53
    ({{
    0.52
     Diffusion
    0.52
     pasta
    0.51
     Doom
    0.51
    াদশ
    0.50
     <!--
    0.50
     Elinor
    0.50
    POSITIVE LOGITS
     ?>
    1.63
    ?>
    1.59
    %>
    1.27
     %>
    1.26
    ?>"
    1.13
    ;?>
    1.12
     ?>"
    1.08
    ()?>
    1.08
     ?></
    1.03
    〉
    0.99
    Activations Density 0.011%

    No Known Activations