© 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. 4322
    Prev
    Next
    INDEX
    Explanations

    phrases related to online subscriptions and support for a news organization

    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
    128
    +0.11
    0.5%
    393
    +0.11
    0.5%
    74
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    474
    +0.11
    0.02
    128
    +0.11
    0.02
    74
    +0.10
    0.02
    Negative Logits
    dfx
    -0.72
    @@
    -0.69
    Standing
    -0.69
    oof
    -0.65
    Ga
    -0.65
    Actually
    -0.64
    tis
    -0.64
    ³³³³³³³³³³³³³³³³
    -0.63
    cffffcc
    -0.62
    adish
    -0.62
    POSITIVE LOGITS
    romeda
    1.04
     subscribe
    0.86
    rea
    0.78
     receive
    0.74
    rew
    0.74
     enjoy
    0.69
     then
    0.65
     learn
    0.64
     customize
    0.64
     explore
    0.63
    Activations Density 0.057%

    No Known Activations