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

    Neuronpedia

    Natural Language
    Autoencoders
    NEW
    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-12B
    3. 24-GEMMASCOPE-2-RES-16K
    4. 11168
    Prev
    Next
    INDEX
    Explanations

    python imports

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

    The neuron fires on programming‐style tokens—things like namespaced identifiers (e.g. “ios::”), compiler flags and settings (e.g. “-O3”, “std=c99”), protocol or API version strings (e.g. “HTTP/1.1”), and library/function calls—essentially spotting code syntax bits rather than natural language.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408

    scope resolution operators or namespace separators in programming code (specifically the "::" symbol).

    oai_token-act-pair · claude-4-5-sonnetTriggered by @jyhe0408

    technical code and web syntax fragments, especially identifiers or paths that include symbols like slashes, dashes, underscores, or double colons (e.g., file paths, URLs, namespaces, flags, and CSS/module names).

    oai_token-act-pair · gpt-5Triggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-12b-pt/resid_post/layer_24_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
    Workflow
    0.36
    しょ
    0.33
    bem
    0.32
    徵
    0.32
    grunt
    0.31
    _
    0.31
    teau
    0.30
    蠔
    0.30
    蚜
    0.30
    bera
    0.30
    POSITIVE LOGITS
     redes
    0.34
     untuk
    0.33
     presupp
    0.31
     پیغمبر
    0.30
     Євро
    0.30
     justific
    0.30
     intensidade
    0.29
     $(`.
    0.29
     retr
    0.29
     ejercicios
    0.29
    Activations Density 0.002%

    No Known Activations