© 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. 13008
    Prev
    Next
    INDEX
    Explanations

    technical information related to computer programming, software bugs, and user interactions in online forums

    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
    +0.18
    0.8%
    255
    +0.11
    0.5%
    6
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    6
    +0.18
    0.10
    640
    +0.11
    0.09
    107
    +0.09
    0.06
    Negative Logits
     credential
    -0.78
    ensable
    -0.76
     unle
    -0.75
    reet
    -0.74
    ère
    -0.74
    intendent
    -0.74
     eleph
    -0.72
     enriched
    -0.71
     enrolled
    -0.71
     privileged
    -0.70
    POSITIVE LOGITS
     Anyway
    1.39
     :(
    1.31
     Seems
    1.25
    Anyway
    1.24
     :)
    1.20
     Logged
    1.16
     ;)
    1.15
     Also
    1.13
    EDIT
    1.13
     Basically
    1.10
    Activations Density 0.370%

    No Known Activations