© 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. 13288
    Prev
    Next
    INDEX
    Explanations

    references to email subscriptions and newsletters

    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
    373
    +0.14
    0.6%
    559
    +0.11
    0.5%
    689
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    559
    +0.14
    0.01
    508
    +0.11
    0.01
    689
    +0.09
    0.01
    Negative Logits
    abouts
    -0.63
    numbered
    -0.62
    rans
    -0.59
    prototype
    -0.58
    bra
    -0.57
    sf
    -0.56
    brush
    -0.53
    ife
    -0.53
    spring
    -0.53
    ball
    -0.53
    POSITIVE LOGITS
    itte
    0.85
    iments
    0.80
    IENCE
    0.79
    ispers
    0.73
     subscriptions
    0.70
    iru
    0.70
    iment
    0.68
     subscribing
    0.67
     subscribe
    0.67
    interstitial
    0.65
    Activations Density 0.007%

    No Known Activations