© 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. 13260
    Prev
    Next
    INDEX
    Explanations

    phrases related to legal processes and judicial actions

    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
    419
    +0.12
    0.5%
    185
    +0.10
    0.5%
    201
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    419
    +0.12
    0.03
    723
    +0.10
    0.02
    39
    +0.10
    0.03
    Negative Logits
    wagen
    -0.70
    Laughs
    -0.68
     TAMADRA
    -0.67
     toler
    -0.65
    KK
    -0.64
     Episcopal
    -0.60
     Zed
    -0.59
     Bomber
    -0.59
    Gas
    -0.59
     chuck
    -0.58
    POSITIVE LOGITS
    processing
    0.97
     analytics
    0.92
    intensive
    0.90
    nsics
    0.88
    abytes
    0.86
     Analytics
    0.83
    transfer
    0.82
    chnology
    0.79
    abyte
    0.76
     formats
    0.73
    Activations Density 0.278%

    No Known Activations