© 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 1536
    9. 8-RES_FS1536-JB
    11. 1123
    Prev
    Next
    INDEX
    Explanations

    phrases related to societal issues and responses to them, particularly pertaining to political, economic, and social aspects

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs1536-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    1,536
    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
    447
    +0.15
    0.7%
    134
    +0.13
    0.6%
    605
    +0.12
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    401
    +0.15
    0.18
    288
    +0.13
    0.21
    527
    +0.12
    0.07
    Negative Logits
     Invention
    -0.71
    foundland
    -0.65
    :,
    -0.64
    ;;;;
    -0.61
     sqor
    -0.59
     noting
    -0.57
    .<
    -0.57
    margin
    -0.57
    rising
    -0.54
    asma
    -0.54
    POSITIVE LOGITS
     could
    1.16
     might
    1.15
     exists
    1.13
     cannot
    1.13
     would
    1.13
     should
    1.07
     existed
    1.07
     couldn
    1.00
     succeeds
    1.00
     went
    1.00
    Activations Density 4.286%

    No Known Activations