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

    Neuronpedia

    INDEX
    Explanations

    acknowledging limitations or inability

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

    role markers that denote the start of the assistant/model’s turn in structured chat transcripts.

    oai_token-act-pair · gpt-5Triggered by @vetterc0
    New Auto-Interp
    Negative Logits
    ście
    0.40
    Salt
    0.39
    ㄤ
    0.37
    ätte
    0.37
    驿
    0.36
     [&
    0.36
     Happens
    0.35
    0.35
    muş
    0.35
    施
    0.34
    POSITIVE LOGITS
     regrett
    0.50
     regret
    0.45
     unfortunately
    0.45
     sorry
    0.43
     عدم
    0.43
     regretted
    0.42
     unable
    0.41
     inability
    0.41
     unavailable
    0.41
     scandalous
    0.40
    Act Density 0.031%

    No Known Activations