© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    1. Home
    3. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 49152
    9. 8-RES_FS49152-JB
    11. 41740
    Prev
    Next
    INDEX
    Explanations

    instances of punctuation and their frequency in the text

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs49152-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    49,152
    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
    393
    +0.13
    0.6%
    357
    +0.12
    0.6%
    100
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    43
    +0.13
    0.02
    384
    +0.12
    0.02
    596
    +0.09
    0.02
    Negative Logits
    enta
    -0.77
    usterity
    -0.75
    onica
    -0.71
    Wire
    -0.69
    verty
    -0.69
    kees
    -0.68
    Track
    -0.68
    aepernick
    -0.67
    igator
    -0.67
    eks
    -0.67
    POSITIVE LOGITS
     vice
    0.77
     whether
    0.70
     namely
    0.68
     somew
    0.68
     how
    0.68
     speculate
    0.63
     succumb
    0.62
     withdrew
    0.61
     congratulate
    0.61
     indulge
    0.60
    Activations Density 0.048%

    No Known Activations