© 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. 18589
    Prev
    Next
    INDEX
    Explanations

    the frequent occurrence of the word "to", specifically in various contexts and verb constructions

    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
    665
    +0.18
    0.8%
    49
    +0.14
    0.6%
    718
    +0.13
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    665
    +0.18
    0.06
    49
    +0.14
    0.05
    718
    +0.13
    0.04
    Negative Logits
    workers
    -0.68
    Lear
    -0.67
    thumbnails
    -0.67
     Explan
    -0.65
     Provided
    -0.64
     Detected
    -0.63
    til
    -0.63
    generated
    -0.63
    valued
    -0.62
    abiding
    -0.61
    POSITIVE LOGITS
     pursue
    1.34
     postpone
    1.28
     dedicate
    1.22
     donate
    1.18
     investigate
    1.17
     ignore
    1.16
     give
    1.16
     revisit
    1.15
     explore
    1.13
     undertake
    1.13
    Activations Density 0.121%

    No Known Activations