© 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. 4725
    Prev
    Next
    INDEX
    Explanations

    prepositions "of" occurring in descriptive phrases

    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
    481
    +0.14
    0.6%
    393
    +0.12
    0.5%
    447
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    608
    +0.14
    0.03
    272
    +0.12
    0.02
    705
    +0.10
    0.02
    Negative Logits
    Sport
    -0.75
    keys
    -0.73
    WF
    -0.72
    enth
    -0.70
    ients
    -0.70
    hesis
    -0.69
    Amount
    -0.69
    Avg
    -0.69
    hen
    -0.68
    herer
    -0.66
    POSITIVE LOGITS
     owning
    1.01
     creating
    0.95
     preserving
    0.93
     eliminating
    0.93
     replacing
    0.91
     needing
    0.90
     restoring
    0.90
     having
    0.88
     extending
    0.88
     removing
    0.87
    Activations Density 0.082%

    No Known Activations