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

    Neuronpedia

    INDEX
    Explanations

    refusing harmful user requests

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Negative Logits
     souci
    0.79
    咱
    0.79
     Approach
    0.78
    んだけど
    0.75
     rethinking
    0.74
    leadsto
    0.74
     ਖ
    0.74
    Approach
    0.72
    機會
    0.71
     dyspe
    0.70
    POSITIVE LOGITS
     functionalities
    0.99
     ideologies
    0.90
     ultimatum
    0.89
     morals
    0.88
     advancements
    0.83
     parameters
    0.83
     persona
    0.82
     ideology
    0.82
     premise
    0.81
     skillset
    0.81
    Act Density 0.179%

    No Known Activations