© 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 12288
    9. 8-RES_FS12288-JB
    11. 1515
    Prev
    Next
    INDEX
    Explanations

    technology-related terms and programming language elements

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs12288-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_12288
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    12,288
    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
    503
    +0.10
    0.4%
    508
    +0.10
    0.4%
    526
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    288
    +0.10
    0.08
    575
    +0.10
    0.06
    678
    +0.09
    0.04
    Negative Logits
    athi
    -0.65
     interviewed
    -0.65
     urge
    -0.63
     surveyed
    -0.61
     resemblance
    -0.58
     advance
    -0.58
    @#&
    -0.57
     Telephone
    -0.56
     contacted
    -0.56
     Bite
    -0.56
    POSITIVE LOGITS
    livion
    1.25
     manageable
    0.84
    orable
    0.84
    Normal
    0.68
    ongh
    0.67
    fledged
    0.67
    oreal
    0.66
     equivalents
    0.65
    olesc
    0.64
    ().
    0.63
    Activations Density 0.749%

    No Known Activations