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

    Neuronpedia

    INDEX
    Explanations

    dark themes like loss

    np_acts-logits-general · gemini-2.5-flash-lite

    contractions and possessives marked by apostrophes.

    oai_token-act-pair · gpt-5Triggered by @yooniel31

    The neuron is picking out slot‐machine jargon and feature headings (e.g. “slots,” “RTP,” “volatility,” “max win,” “bonus features,” etc.).

    oai_token-act-pair · o4-miniTriggered by @yooniel31
    New Auto-Interp
    Negative Logits
    an
    0.79
    a
    0.66
    er
    0.64
    ar
    0.63
    z
    0.62
    u
    0.61
    на
    0.59
    in
    0.57
    ↵
    0.57
    ed
    0.57
    POSITIVE LOGITS
    0.69
     is
    0.59
     are
    0.59
     in
    0.53
     an
    0.50
     to
    0.49
     a
    0.49
     è
    0.48
     as
    0.47
    ،
    0.47
    Act Density 15.937%

    No Known Activations