© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it
    Neuronpedia
    Jacobian LensNEW
    Natural Language
    Autoencoders
    MAY 26    Assistant AxisJAN 26Circuit TracerAPR 26SteerSAE EvalsExportsinterp-engineNEW Community BlogPrivacy & TermsContact
    1. Home
    3. GPT2-Small
    5. Transcoders Residuals
    7. 8-TRES-DC
    9. 3292
    Prev
    Next
    INDEX
    Explanations

    instances of the word "were" and related statistics or comparisons

    oai_token-act-pair · gpt-4o-mini
    Triggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Embeds
    IFrame
    Link
    Negative Logits
    was
    -0.76
    FTWARE
    -0.71
    iosyncr
    -0.71
    could
    -0.68
     underwent
    -0.67
    cape
    -0.66
     WAS
    -0.65
    ukemia
    -0.65
     Lauder
    -0.62
     Doct
    -0.62
    POSITIVE LOGITS
     happier
    0.86
    ��
    0.71
    �士
    0.70
     characterized
    0.70
     somewhere
    0.69
    gin
    0.67
    �
    0.67
    ��
    0.64
     majority
    0.64
     lighter
    0.64
    Activations Density 0.257%

    No Known Activations