© 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. 5383
    Prev
    Next
    INDEX
    Explanations

    insights or analysis in various contexts, such as research, politics, or personal interactions

    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
    634
    +0.12
    0.6%
    195
    +0.12
    0.5%
    744
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    634
    +0.12
    0.02
    744
    +0.12
    0.01
    399
    +0.11
    0.01
    Negative Logits
    ategory
    -0.70
    senal
    -0.63
     Vengeance
    -0.60
    emale
    -0.59
    ouf
    -0.57
     Roses
    -0.57
     fuss
    -0.56
     disbanded
    -0.56
    HAM
    -0.55
     Laurent
    -0.55
    POSITIVE LOGITS
    ibility
    1.10
     glean
    1.02
     insights
    0.98
     into
    0.98
     insight
    0.95
    ively
    0.94
    ibly
    0.93
    ible
    0.90
    spection
    0.89
     INTO
    0.87
    Activations Density 0.034%

    No Known Activations