© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    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. 12158
    Prev
    Next
    INDEX
    Explanations

    information related to software updates and technology releases

    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
    640
    +0.23
    1.0%
    679
    +0.14
    0.7%
    120
    +0.13
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    6
    +0.23
    0.14
    679
    +0.14
    0.12
    640
    +0.13
    0.11
    Negative Logits
     occupation
    -0.86
     occupations
    -0.82
     involuntary
    -0.81
     volunteer
    -0.79
     imperson
    -0.78
     lifes
    -0.77
     discipline
    -0.77
     contam
    -0.76
     concentration
    -0.76
     intervention
    -0.76
    POSITIVE LOGITS
     Interestingly
    1.35
     Additionally
    1.34
     Thankfully
    1.29
     However
    1.28
     Considering
    1.27
     Apparently
    1.26
     Notably
    1.26
     Regardless
    1.25
     According
    1.25
     Hopefully
    1.24
    Activations Density 0.560%

    No Known Activations