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

    Neuronpedia

    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. 2218
    Prev
    Next
    INDEX
    Explanations

    information about incidents involving immigrants and possibly illegal activities

    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
    +0.20
    0.9%
    739
    +0.12
    0.5%
    683
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    373
    +0.20
    -0.00
    683
    +0.12
    0.01
    739
    +0.10
    0.01
    Negative Logits
    adal
    -0.67
    nea
    -0.67
    naire
    -0.66
    amacare
    -0.65
    lain
    -0.64
    adian
    -0.63
    ĪĴ
    -0.63
    gling
    -0.63
    acea
    -0.63
    aba
    -0.62
    POSITIVE LOGITS
     IMAGES
    2.08
     Images
    0.95
     SPORTS
    0.94
     PHOTO
    0.84
    URE
    0.83
    MORE
    0.79
    MEN
    0.79
    Images
    0.78
     POLIT
    0.77
     WATCH
    0.77
    Activations Density 0.013%

    No Known Activations