© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 24576
    9. 8-RES_FS24576-JB
    11. 16428
    Prev
    Next
    INDEX
    Explanations

    phrases related to political figures and their actions or comments

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs24576-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    24,576
    Data Type
    torch.float32
    Hook Point
    blocks.8.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    8
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    720
    +0.12
    0.5%
    61
    +0.12
    0.5%
    447
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    472
    +0.12
    0.04
    720
    +0.12
    0.05
    61
    +0.10
    0.04
    Negative Logits
    emis
    -0.79
    neau
    -0.78
    inct
    -0.74
    allery
    -0.70
     batch
    -0.69
    folio
    -0.68
    burgh
    -0.67
     Printed
    -0.67
    fen
    -0.67
    raq
    -0.66
    POSITIVE LOGITS
     Mexicans
    1.05
     homosexuals
    0.94
     gays
    0.93
     insulting
    0.90
     himself
    0.86
     hating
    0.83
     derog
    0.83
     Muslims
    0.83
     dispar
    0.81
     homosexuality
    0.81
    Activations Density 0.364%

    No Known Activations