© 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. 13366
    Prev
    Next
    INDEX
    Explanations

    terms related to paperwork and bureaucracy

    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
    447
    +0.11
    0.5%
    221
    +0.09
    0.4%
    404
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    326
    +0.11
    0.04
    99
    +0.09
    0.04
    404
    +0.09
    0.04
    Negative Logits
    ERG
    -0.83
    CVE
    -0.74
    ailability
    -0.74
    ei
    -0.74
    İĭ
    -0.74
    IAS
    -0.74
    EW
    -0.73
    ¿
    -0.73
    unci
    -0.73
    itia
    -0.72
    POSITIVE LOGITS
     Soup
    1.08
     Circus
    1.02
     Pie
    0.99
     Girl
    0.97
     Grove
    0.95
     Connection
    0.95
     Madness
    0.94
     Alley
    0.93
     Massacre
    0.92
     Boys
    0.91
    Activations Density 0.220%

    No Known Activations