© 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. 5250
    Prev
    Next
    INDEX
    Explanations

    names of specific individuals mixed with legal terms or actions

    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
    480
    +0.32
    1.6%
    481
    +0.15
    0.7%
    430
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    232
    +0.32
    0.10
    659
    +0.15
    0.08
    367
    +0.10
    0.08
    Negative Logits
    staking
    -0.81
    mble
    -0.78
     pse
    -0.74
    cellaneous
    -0.68
     irony
    -0.65
    lished
    -0.64
     narrated
    -0.62
    ographed
    -0.62
     puzz
    -0.61
     dashed
    -0.61
    POSITIVE LOGITS
    addafi
    0.80
     Bashar
    0.77
    ullah
    0.76
     Tsarnaev
    0.72
    xit
    0.72
    ratom
    0.70
    ghazi
    0.70
    odan
    0.68
    inia
    0.68
    SPA
    0.66
    Activations Density 0.612%

    No Known Activations