© 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. 129347
    Prev
    Next
    INDEX
    Explanations

    The neuron detects imperative‐mood verbs—that is, command forms like “notice,” “sneak,” “ask,” “pass,” “while,” etc., especially when addressing “you.”

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

    hidden or concealed items

    np_acts-logits-general · gemini-2.5-flash-lite
    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
     before
    -1.02
     even
    -0.99
     despite
    -0.96
     sixteen
    -0.93
     lively
    -0.93
     these
    -0.91
     every
    -0.89
     there
    -0.88
     thirteen
    -0.88
     this
    -0.86
    POSITIVE LOGITS
     sly
    1.57
     stealth
    1.52
     secretly
    1.51
     quietly
    1.34
     secret
    1.30
     discreet
    1.29
    偷偷
    1.29
    secret
    1.29
    hidden
    1.28
    実は
    1.26
    Activations Density 0.039%

    No Known Activations