© 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 EvalsExportsAPI 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. 33
    Prev
    Next
    INDEX
    Explanations

    references to political and governmental issues

    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
    480
    +0.48
    2.7%
    393
    +0.20
    1.1%
    362
    +0.17
    1.0%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    362
    +0.48
    0.20
    608
    +0.20
    0.15
    393
    +0.17
    0.14
    Negative Logits
     besides
    -0.67
    tackle
    -0.64
    !!!
    -0.63
    !!!!
    -0.62
     because
    -0.61
    !!!!!
    -0.60
     wherein
    -0.59
     namely
    -0.59
    aurus
    -0.58
     patiently
    -0.57
    POSITIVE LOGITS
     latter
    1.30
     same
    1.10
     slightest
    1.00
     aforementioned
    0.99
     remainder
    0.91
     country
    0.89
     latest
    0.88
     ensuing
    0.88
     entirety
    0.88
     nation
    0.87
    Activations Density 0.959%

    No Known Activations