© 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. 1828
    Prev
    Next
    INDEX
    Explanations

    instances of the word "too."

    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
    132
    +0.13
    0.6%
    745
    +0.11
    0.5%
    55
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    132
    +0.13
    0.03
    55
    +0.11
    0.03
    513
    +0.10
    0.03
    Negative Logits
     Hammond
    -0.62
    farious
    -0.61
    estamp
    -0.60
    è¦ļéĨĴ
    -0.59
    Rus
    -0.58
     Slash
    -0.58
    76561
    -0.57
    inators
    -0.57
    ixel
    -0.57
    emort
    -0.56
    POSITIVE LOGITS
     beware
    0.78
    leep
    0.75
     afraid
    0.73
    oths
    0.69
    ille
    0.68
     reliant
    0.67
    urden
    0.67
     busy
    0.67
    ISTORY
    0.66
    nder
    0.64
    Activations Density 0.025%

    No Known Activations