© 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. 14398
    Prev
    Next
    INDEX
    Explanations

    locations related to specific events or individuals

    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
    757
    +0.12
    0.5%
    98
    +0.11
    0.5%
    764
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    232
    +0.12
    0.05
    757
    +0.11
    0.04
    530
    +0.11
    0.04
    Negative Logits
     Clim
    -0.73
     mete
    -0.72
     bottled
    -0.64
    20439
    -0.63
     intr
    -0.62
     penet
    -0.62
     decom
    -0.61
     Rover
    -0.59
     Hubble
    -0.59
     isot
    -0.59
    POSITIVE LOGITS
    burgh
    1.02
     Janeiro
    0.94
     Rapids
    0.91
    opolis
    0.88
     City
    0.88
    town
    0.83
    nesota
    0.82
    ottesville
    0.82
    hots
    0.81
    shire
    0.81
    Activations Density 0.134%

    No Known Activations