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

    Neuronpedia

    INDEX
    Explanations

    measurements/comparisons

    np_max-act · gemini-2.0-flash

    This neuron detects comparative adjectives or adverbs indicating a relative “more/less” quality (e.g., “leichter,” “weicher”).

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
     happier
    -0.08
     lower
    -0.08
     larger
    -0.07
     better
    -0.07
     more
    -0.07
     less
    -0.07
    24
    -0.07
    %)↵↵
    -0.07
     nicer
    -0.07
     easier
    -0.07
    POSITIVE LOGITS
    "><?
    0.06
     getObject
    0.06
     Pg
    0.06
     прек
    0.06
     disposit
    0.06
    (key
    0.06
     hard
    0.06
     گذ
    0.06
     potřeb
    0.06
     belir
    0.06
    Act Density 0.152%

    No Known Activations