© 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 24576
    9. 8-RES_FS24576-JB
    11. 7305
    Prev
    Next
    INDEX
    Explanations

    technical terms related to software and technology

    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
    87
    +0.34
    1.7%
    481
    +0.14
    0.7%
    459
    +0.13
    0.7%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    87
    +0.34
    -0.03
    459
    +0.14
    0.09
    575
    +0.13
    0.09
    Negative Logits
    ewitness
    -0.57
    wcsstore
    -0.56
    untarily
    -0.55
    catentry
    -0.52
     Publication
    -0.52
    )?
    -0.52
    pires
    -0.52
    NSA
    -0.51
     Represent
    -0.50
     Answers
    -0.50
    POSITIVE LOGITS
    ;
    0.93
     because
    0.90
     though
    0.90
     whereas
    0.87
     but
    0.84
     although
    0.80
     however
    0.75
    .;
    0.73
    %;
    0.71
     unless
    0.69
    Activations Density 1.591%

    No Known Activations