© 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. 14140
    Prev
    Next
    INDEX
    Explanations

    words related to epitomizing or embodying a concept or characteristic

    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
    85
    +0.13
    0.6%
    386
    +0.12
    0.5%
    276
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    85
    +0.13
    0.01
    276
    +0.12
    0.01
    199
    +0.11
    0.01
    Negative Logits
     Worldwide
    -0.71
    WARD
    -0.71
    ITED
    -0.70
    BSD
    -0.67
     Panther
    -0.64
    SOURCE
    -0.62
    theless
    -0.62
     Safari
    -0.59
     Wonderland
    -0.59
     Bucks
    -0.59
    POSITIVE LOGITS
    hetically
    1.27
    helial
    1.27
    hetical
    1.24
    hetic
    1.16
    ographs
    1.16
    ographical
    1.09
    ograph
    1.09
    ographies
    1.07
    hesis
    1.07
    orical
    1.06
    Activations Density 0.019%

    No Known Activations