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

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsinterp-engineNEWAPI 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. 2636
    Prev
    Next
    INDEX
    Explanations

    mentions of significant numerical quantities

    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
    34
    +0.08
    0.5%
    459
    +0.07
    0.5%
    298
    +0.07
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    298
    +0.08
    0.05
    620
    +0.07
    0.01
    460
    +0.07
    0.04
    Negative Logits
    ActionCode
    -0.57
    Ą
    -0.56
    ė
    -0.56
    ãĤ¤
    -0.56
    ý
    -0.56
    -0.56
    ÷
    -0.56
     RandomRedditor
    -0.56
     TheNitrome
    -0.56
    Ĉ
    -0.56
    POSITIVE LOGITS
     respectively
    0.59
    .''.
    0.48
     Lans
    0.47
    '.
    0.45
     CrossRef
    0.44
    .</
    0.44
     Aven
    0.43
    .
    0.43
     curfew
    0.43
    .).
    0.42
    Activations Density 1.586%

    No Known Activations