© 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-2-27B
    5. 22-GEMMASCOPE-RES-131K
    7. 65823
    Prev
    Next
    INDEX
    Explanations

    links and structural markers

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

    The neuron strongly activates on left‐square‐bracket tokens that introduce numeric metadata (e.g. “[ 1064”, “[ 1048”, etc.), i.e. bracketed reference or page markers.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_22/width_131k
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    は
    -1.73
    があります
    -1.53
     has
    -1.45
    を
    -1.41
    である
    -1.41
    や
    -1.38
    で
    -1.34
    <bos>
    -1.34
     is
    -1.32
    それは
    -1.32
    POSITIVE LOGITS
    総評
    1.62
    為に
    1.57
     >",
    1.42
    の方に
    1.38
    の方法
    1.37
     découver
    1.36
    ","\
    1.36
    に使用
    1.35
     preocupar
    1.35
    が可能
    1.34
    Activations Density 0.025%

    No Known Activations