© 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. 2757
    Prev
    Next
    INDEX
    Explanations

    information related to a specific disease, in this case, Ebola

    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
    87
    +0.33
    1.6%
    470
    +0.11
    0.6%
    420
    +0.09
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    87
    +0.33
    -0.02
    470
    +0.11
    0.04
    157
    +0.09
    0.06
    Negative Logits
    ."[
    -0.67
    SPONSORED
    -0.62
     accordingly
    -0.62
    âĢ¦]
    -0.61
     [...]
    -0.61
     thereof
    -0.61
    ".[
    -0.61
    Zip
    -0.60
     sanity
    -0.59
    ).[
    -0.59
    POSITIVE LOGITS
     controvers
    0.86
     notoriously
    0.81
    quartered
    0.78
     famously
    0.76
     controversial
    0.73
     plagued
    0.73
     fiercely
    0.69
     accused
    0.68
     dogged
    0.67
     traditionally
    0.66
    Activations Density 1.135%

    No Known Activations