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

    Neuronpedia

    INDEX
    Explanations

    leading to negative consequences

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Negative Logits
     slightly
    0.61
     somewhat
    0.59
     Slightly
    0.59
     légèrement
    0.57
     slight
    0.56
     agak
    0.56
     Certain
    0.56
     particular
    0.55
     particolari
    0.55
     Slight
    0.55
    POSITIVE LOGITS
     unpredictable
    0.78
     worse
    0.68
     disaster
    0.68
     unpredict
    0.68
     seriously
    0.66
     disastrous
    0.66
     meaningless
    0.65
     useless
    0.64
     unreliable
    0.64
     Worse
    0.64
    Act Density 0.038%

    No Known Activations