© 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. 4325
    Prev
    Next
    INDEX
    Explanations

    statements related to social commentary and opinions on various societal matters

    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
    450
    +0.10
    0.5%
    447
    +0.10
    0.4%
    732
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    582
    +0.10
    0.02
    715
    +0.10
    0.03
    739
    +0.10
    0.01
    Negative Logits
    Interstitial
    -0.69
     otherwise
    -0.66
    istani
    -0.65
    tones
    -0.61
    Vert
    -0.61
     Newspaper
    -0.61
    units
    -0.58
    shown
    -0.57
    ifled
    -0.57
    angering
    -0.57
    POSITIVE LOGITS
    aucus
    0.78
     begs
    0.75
    oiler
    0.68
    bye
    0.68
     wondering
    0.67
     congratulations
    0.67
     concludes
    0.67
     parting
    0.67
     yeah
    0.67
     caution
    0.67
    Activations Density 0.260%

    No Known Activations