Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    INDEX
    Explanations

    Comparisons and contrasts

    np_max-act · gemini-2.0-flash

    This neuron detects discourse‐level transitional or contrastive adverbs and cue words (e.g. “instead,” “contrary,” “however”) that signal shifts or contrasts in the text.

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
    ecessary
    -0.07
    assel
    -0.07
    abe
    -0.07
     owing
    -0.06
    搜
    -0.06
     âm
    -0.06
    ollower
    -0.06
    .resize
    -0.06
    ेब
    -0.06
     overtime
    -0.06
    POSITIVE LOGITS
     desperation
    0.06
     record
    0.06
    ющей
    0.06
     primaryKey
    0.06
    ाखण
    0.06
    =target
    0.06
    .char
    0.06
     CPI
    0.06
     imagined
    0.06
    =>'
    0.06
    Act Density 0.114%

    No Known Activations