© 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. 22914
    Prev
    Next
    INDEX
    Explanations

    key statistical and analytical phrases related to societal issues and the effects of interventions

    oai_token-act-pair · gpt-4o-miniTriggered 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
    87
    +0.32
    1.6%
    447
    +0.16
    0.9%
    481
    +0.14
    0.7%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    288
    +0.32
    0.10
    401
    +0.16
    0.08
    87
    +0.14
    -0.02
    Negative Logits
    olulu
    -0.66
    !.
    -0.58
     Invention
    -0.57
    .<
    -0.56
    idav
    -0.55
    };
    -0.52
    :,
    -0.52
     sqor
    -0.52
    cellaneous
    -0.51
    SourceFile
    -0.51
    POSITIVE LOGITS
     should
    0.91
     could
    0.88
     exists
    0.83
     might
    0.83
     ought
    0.82
     cannot
    0.81
     shouldn
    0.80
     would
    0.79
     existed
    0.79
     SHOULD
    0.78
    Activations Density 0.840%

    No Known Activations