© 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. 5512
    Prev
    Next
    INDEX
    Explanations

    mentions of various people in the context of interviews, opinions, and roles in sports and politics

    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
    51
    +0.11
    0.5%
    233
    +0.10
    0.5%
    321
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    659
    +0.11
    0.06
    232
    +0.10
    0.04
    545
    +0.10
    0.03
    Negative Logits
    ça
    -0.71
     sorcery
    -0.63
    ACTED
    -0.61
    phrine
    -0.61
     Dak
    -0.57
    JUST
    -0.57
     Cosby
    -0.57
    ruary
    -0.57
     linen
    -0.57
     orphans
    -0.57
    POSITIVE LOGITS
    angan
    0.84
    argon
    0.82
    lear
    0.68
    Ĥ¬
    0.67
    ĸļ
    0.67
    boxing
    0.67
    esy
    0.66
    yden
    0.65
     Lumpur
    0.64
    strom
    0.64
    Activations Density 0.106%

    No Known Activations