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

    Neuronpedia

    INDEX
    Explanations

    Spelling errors or multiple languages

    np_max-act · gemini-2.0-flash

    The neuron spikes on tokens containing Polish letters with diacritics (ą, ć, ę, ł, ń, ś, ó, ż, ź).

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8

    words that contain specific Polish letters.

    oai_token-act-pair · gpt-4o-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
     tapes
    -0.07
     ры
    -0.07
     tape
    -0.06
     contributes
    -0.06
     environments
    -0.06
     -------------------------------------------------------------------------
    -0.06
     masking
    -0.06
     accidentally
    -0.06
     prosecution
    -0.06
    Presentation
    -0.06
    POSITIVE LOGITS
    exampleModalLabel
    0.07
    .imag
    0.07
     BaseService
    0.06
     asylum
    0.06
    ━━━━━━━━
    0.06
    額
    0.06
     vigor
    0.06
    .Com
    0.06
     모습
    0.06
     Malik
    0.06
    Act Density 0.015%

    No Known Activations