© 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. 20555
    Prev
    Next
    INDEX
    Explanations

    phrases related to legal and political discussions

    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
    420
    +0.15
    0.7%
    447
    +0.14
    0.6%
    411
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    121
    +0.15
    0.05
    647
    +0.14
    0.04
    264
    +0.10
    0.03
    Negative Logits
    ahime
    -0.91
    bis
    -0.91
    ramid
    -0.80
     guiActiveUnfocused
    -0.73
    ensitivity
    -0.72
    ens
    -0.71
    scale
    -0.69
    format
    -0.69
     decay
    -0.68
    lag
    -0.68
    POSITIVE LOGITS
     representatives
    1.16
     former
    1.03
     Reps
    0.94
     fellow
    0.93
     Mohamed
    0.92
     Mohammad
    0.92
     Abd
    0.90
     members
    0.89
     retired
    0.88
     Christine
    0.88
    Activations Density 0.238%

    No Known Activations