© 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
    2. Gemma-3-12B
    3. 31-GEMMASCOPE-2-RES-16K
    4. 12744
    Prev
    Next
    INDEX
    Explanations

    select ' ( .` jq 'This neuron appears to be activated by patterns related to JSON manipulation, specifically using the `jq` command and its syntax within code or scripting contexts. The tokens like `jq`, `'`, `.`, `[`, `(`, and `select` are all common elements in `jq` queries used for filtering and transforming JSON data. The `TOP_POSITIVE_LOGITS` also contain characters like `[@`, `(@`, `//@`, `//*[@`, which often appear in contexts where structured data like JSON or XML is being parsed, similar to `jq`'s usage.Therefore, a concise explanation capturing this behavior would be related to `jq` syntax.Explanation:`jq` query syntax

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

    The neuron detects XPath‐style navigation and selection vocabulary—tokens used for traversing XML node hierarchies (e.g. “select,” “parent,” “using,” “..,” “//,” “@,” “|”).

    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
     gracia
    0.53
     及
    0.53
     llevará
    0.52
     Designed
    0.52
    Designed
    0.51
     اچھا
    0.51
    iality
    0.50
     बहा
    0.50
    hills
    0.50
     designed
    0.50
    POSITIVE LOGITS
    [@
    1.05
     [@
    0.79
    normalize
    0.75
     normalize
    0.74
    //@
    0.68
    /@
    0.64
    //*[@
    0.64
    (@
    0.62
     //@
    0.61
     (@
    0.60
    Activations Density 0.004%

    No Known Activations