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

    Neuronpedia

    INDEX
    Explanations

    Failure of AI or code

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

    assistant/model responses that provide structured explanations or evaluations—especially noting flaws, limitations, or following task instructions.

    oai_token-act-pair · gpt-5Triggered by @irajjmoradi
    New Auto-Interp
    Negative Logits
     positive
    0.72
     enrich
    0.69
     enriching
    0.69
     stär
    0.69
    enrich
    0.68
     exhilarating
    0.66
     Enh
    0.66
     favorable
    0.65
     enhancing
    0.65
    喜爱
    0.64
    POSITIVE LOGITS
     useless
    1.72
     failed
    1.68
     ineffective
    1.64
    failed
    1.59
     incapable
    1.55
     futile
    1.53
     pointless
    1.53
     Failed
    1.50
     worthless
    1.49
     unsuccessful
    1.48
    Act Density 3.409%

    No Known Activations