© 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. 16416
    Prev
    Next
    INDEX
    Explanations

    names of individuals or organizations related to various topics

    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
    659
    +0.11
    0.5%
    314
    +0.11
    0.5%
    409
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    659
    +0.11
    0.04
    574
    +0.11
    0.03
    409
    +0.10
    0.03
    Negative Logits
    grass
    -0.73
     removable
    -0.68
     Schumer
    -0.66
     Nieto
    -0.66
     slack
    -0.66
     Granger
    -0.64
     CPR
    -0.64
     reversible
    -0.63
     limp
    -0.62
     Totem
    -0.62
    POSITIVE LOGITS
    ICH
    1.26
    senal
    1.23
    ansom
    1.23
    haps
    1.22
    OUND
    1.19
    ounding
    1.19
    ascal
    1.16
    outine
    1.15
    outing
    1.15
    alph
    1.14
    Activations Density 0.041%

    No Known Activations