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

    Neuronpedia

    INDEX
    Explanations

    neural network code

    np_max-act · gemini-2.0-flash

    The neuron strongly activates on tokens that occur inside the assistant’s enumerated, step‐by‐step or list‐style instructions (e.g. “Day 1: …,” “Round 2: …,” bullet‐point steps), and remains silent on plain user or system text.

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
     солн
    -0.07
    OutOfBoundsException
    -0.07
    ุญ
    -0.06
    自身
    -0.06
     terrestrial
    -0.06
     ler
    -0.06
    Cheers
    -0.06
     게
    -0.06
     Pron
    -0.06
    ционной
    -0.06
    POSITIVE LOGITS
    urable
    0.07
    dbName
    0.07
     Cognitive
    0.06
    .Gen
    0.06
    _sector
    0.06
    >I
    0.06
    struction
    0.06
    wall
    0.06
     patent
    0.06
    .CH
    0.06
    Act Density 0.005%

    No Known Activations