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

    Neuronpedia

    INDEX
    Explanations

    common phrases like "there's one thing" or "one thing that" expressing a singular point or issue

    oai_token-act-pair · gpt-3.5-turbo

    constructions indicating conditional statements or hypothetical scenarios

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Negative Logits
    çīĪ
    -0.72
    ãĤ¶
    -0.69
    ãĥķ
    -0.67
     Estimates
    -0.63
    éĹ
    -0.63
    "],
    -0.62
    Ì
    -0.62
    bart
    -0.60
    jab
    -0.60
     Som
    -0.58
    POSITIVE LOGITS
     sufficiently
    0.71
     indeed
    0.70
     truly
    0.70
     something
    0.69
    nt
    0.68
     properly
    0.68
     REALLY
    0.65
     somehow
    0.65
    oubted
    0.64
    acle
    0.62
    Act Density 0.226%

    No Known Activations