© 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. 4181
    Prev
    Next
    INDEX
    Explanations

    controversial and legal topics, such as legislation, court cases, and political statements

    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
    447
    +0.44
    2.4%
    480
    +0.35
    1.9%
    481
    +0.26
    1.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    288
    +0.44
    0.13
    481
    +0.35
    0.10
    575
    +0.26
    0.10
    Negative Logits
    »Ĵ
    -0.61
    lished
    -0.58
    idav
    -0.53
    rocal
    -0.50
    mble
    -0.48
     Administ
    -0.48
     UM
    -0.47
    OTAL
    -0.47
    Detailed
    -0.47
    erenn
    -0.46
    POSITIVE LOGITS
     anymore
    0.74
     instead
    0.64
     inappropriately
    0.61
     amid
    0.57
     altogether
    0.57
     Instead
    0.56
     Belfast
    0.56
     pedoph
    0.55
     unfairly
    0.55
     lest
    0.54
    Activations Density 1.002%

    No Known Activations