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

    Neuronpedia

    INDEX
    Explanations

    types and classes

    np_max-act · gemini-2.0-flash

    The neuron detects tokens that name subdivisions or categories prefixed by “sub,” such as “subclass,” “subtype,” or “subcategories.”

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

    references to subsystems within complex systems.

    oai_token-act-pair · gpt-4o-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
     EOS
    -0.07
     Joshua
    -0.07
     candid
    -0.07
     eos
    -0.06
     minden
    -0.06
     floats
    -0.06
     yere
    -0.06
    .set
    -0.06
     cosy
    -0.06
    	fprintf
    -0.06
    POSITIVE LOGITS
    [sub
    0.09
     subsection
    0.09
     subgroup
    0.09
    subclass
    0.08
     preseason
    0.08
     subtype
    0.08
     subnet
    0.08
     subrange
    0.08
     HG
    0.08
     subtree
    0.08
    Act Density 0.033%

    No Known Activations