© 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 EvalsExportsAPI 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. 6181
    Prev
    Next
    INDEX
    Explanations

    words related to famous individuals named "Frank"

    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
    55
    +0.11
    0.5%
    682
    +0.11
    0.5%
    339
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    682
    +0.11
    0.02
    55
    +0.11
    0.02
    81
    +0.09
    0.02
    Negative Logits
    ILA
    -0.73
    ADRA
    -0.67
    Trend
    -0.66
    resent
    -0.65
    wrapper
    -0.65
    TPPStreamerBot
    -0.63
    halla
    -0.62
    atform
    -0.62
     tampering
    -0.60
     Attention
    -0.60
    POSITIVE LOGITS
    fur
    1.41
    furt
    1.40
    lins
    1.14
    fort
    1.01
    enstein
    0.99
     Geh
    0.92
     Olson
    0.92
    lin
    0.91
     Lloyd
    0.79
    ovi
    0.79
    Activations Density 0.014%

    No Known Activations