© 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 EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-2-27B
    3. 22-GEMMASCOPE-RES-131K
    4. 38433
    Prev
    Next
    INDEX
    Explanations

    interesting and words that follow

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

    The neuron fires strongly on tokens and punctuation used in online chat or forum posts—things like double‐dashes (“--”), live/​stlsa tags (“<live>”), and interjections or user‐to‐user remarks (“interesting,” “You are missing a space”)—so it’s detecting bits of internet‐style message markup and commentary.

    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
    sterreich
    -0.89
     exercer
    -0.88
    ifié
    -0.88
     pesso
    -0.85
    mphs
    -0.85
    ASED
    -0.84
     trás
    -0.83
    itom
    -0.83
     DONALD
    -0.83
     rádio
    -0.82
    POSITIVE LOGITS
     spelling
    0.98
     reconocida
    0.86
     spell
    0.81
     Sahne
    0.81
    뚜
    0.81
     spelled
    0.80
     white
    0.79
    巨大
    0.78
     invitados
    0.77
     as
    0.76
    Activations Density 0.078%

    No Known Activations