© 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. 22923
    Prev
    Next
    INDEX
    Explanations

    references to code structure and syntax in programming languages

    oai_token-act-pair · gpt-4o-miniTriggered 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
    640
    +0.16
    0.7%
    255
    +0.15
    0.7%
    314
    +0.14
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    640
    +0.16
    0.05
    160
    +0.15
    0.04
    442
    +0.14
    0.03
    Negative Logits
     puberty
    -0.67
     Bunker
    -0.65
     Cre
    -0.64
     Carlton
    -0.63
     slam
    -0.63
     royal
    -0.61
     plush
    -0.61
     Canter
    -0.60
     bunk
    -0.60
     lou
    -0.59
    POSITIVE LOGITS
    Notice
    1.20
    Alternatively
    1.19
    Output
    1.19
    Then
    1.06
    then
    1.04
    Note
    1.04
    Example
    1.03
    Similarly
    1.02
    Again
    1.02
    where
    1.02
    Activations Density 0.100%

    No Known Activations