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

    Neuronpedia

    INDEX
    Explanations

    contrastive conjunctions

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

    the neuron detects words and tokens related to military conflict and battle (e.g., battle types, troops, campaigns, and combat actions).

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

    This neuron is looking for numeric information (tokens representing numbers, measurements, or counts).

    oai_token-act-pair · o4-miniTriggered by @yooniel31

    the word "soldiers" or references to military personnel/combatants.

    oai_token-act-pair · claude-4-5-sonnetTriggered by @yooniel31
    New Auto-Interp
    Negative Logits
     Итак
    0.55
     കൊണ്ട്
    0.54
    Итак
    0.49
     ossia
    0.46
    于是
    0.44
     Porque
    0.44
    കൊണ്ട്
    0.44
     Öncelikle
    0.43
    つまり
    0.42
    รษฐ
    0.42
    POSITIVE LOGITS
     although
    1.17
    although
    1.11
     있으며
    0.97
    ，
    0.96
    ,
    0.93
     aunque
    0.89
    ،
    0.89
     though
    0.88
     but
    0.87
    though
    0.87
    Act Density 0.668%

    No Known Activations