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

    Neuronpedia

    INDEX
    Explanations

    by

    np_max-act · gemini-2.0-flash

    The neuron activates on tokens that introduce procedural instructions or advice—especially words like “start,” “by,” and “asking” that cue a step or method.

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
     आपक
    -0.07
     ditch
    -0.07
     kết
    -0.06
    jsx
    -0.06
     اپ
    -0.06
    uchs
    -0.06
     příst
    -0.06
     эф
    -0.06
    �
    -0.06
     grandson
    -0.06
    POSITIVE LOGITS
    pleted
    0.07
     flawed
    0.07
    Republicans
    0.06
    لي
    0.06
    aña
    0.06
     SID
    0.06
     FAT
    0.06
    vertis
    0.06
     consulted
    0.06
    .op
    0.06
    Act Density 0.028%

    No Known Activations