© 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 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. 12397
    Prev
    Next
    INDEX
    Explanations

    statements related to specific statistics or numbers

    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
    355
    +0.13
    0.6%
    757
    +0.11
    0.5%
    745
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    402
    +0.13
    0.05
    637
    +0.11
    0.04
    355
    +0.10
    0.04
    Negative Logits
    Redditor
    -0.64
    inx
    -0.61
    heed
    -0.61
    potion
    -0.61
    paio
    -0.60
    utonium
    -0.59
    inders
    -0.59
    ÃįÃį
    -0.58
    iversary
    -0.58
     Fenrir
    -0.57
    POSITIVE LOGITS
     cases
    1.30
     contexts
    1.26
     respects
    1.25
     situations
    1.07
     areas
    1.07
     case
    1.01
     instances
    1.00
     directions
    0.99
     places
    0.98
     vicinity
    0.98
    Activations Density 0.189%

    No Known Activations