© 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. 17047
    Prev
    Next
    INDEX
    Explanations

    words related to physical objects or written text, specifically emphasizing letters

    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
    707
    +0.12
    0.6%
    285
    +0.11
    0.5%
    390
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    707
    +0.12
    0.02
    234
    +0.11
    0.02
    140
    +0.10
    0.02
    Negative Logits
    rolet
    -0.89
    rises
    -0.72
    eger
    -0.72
    orsi
    -0.72
    roxy
    -0.71
    DERR
    -0.67
    gpu
    -0.67
    ESCO
    -0.67
    ivals
    -0.65
    asio
    -0.65
    POSITIVE LOGITS
    sworth
    0.98
    press
    0.92
    forms
    0.91
    letter
    0.85
     letters
    0.84
    mith
    0.83
     alphabet
    0.81
    boxing
    0.78
    velop
    0.78
    Letter
    0.76
    Activations Density 0.016%

    No Known Activations