© 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. 19657
    Prev
    Next
    INDEX
    Explanations

    the term "key" or related keywords indicating importance in various contexts

    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
    718
    +0.10
    0.5%
    365
    +0.10
    0.4%
    559
    +0.08
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    623
    +0.10
    0.03
    8
    +0.10
    0.04
    381
    +0.08
    0.04
    Negative Logits
    bows
    -0.83
    ancies
    -0.81
    classes
    -0.78
    events
    -0.78
    apses
    -0.77
     curls
    -0.77
     Robots
    -0.76
    atism
    -0.74
    iencies
    -0.74
    onies
    -0.73
    POSITIVE LOGITS
     contributor
    1.24
     indicator
    1.21
     reminder
    1.16
     contender
    1.13
     deterrent
    1.12
     source
    1.09
     example
    1.07
     indication
    1.06
     predictor
    1.06
     motiv
    1.04
    Activations Density 0.091%

    No Known Activations