© 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. 16964
    Prev
    Next
    INDEX
    Explanations

    references to locations or organizations

    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
    481
    +0.13
    0.6%
    40
    +0.10
    0.4%
    151
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    575
    +0.13
    0.08
    40
    +0.10
    0.06
    288
    +0.09
    0.08
    Negative Logits
    htaking
    -0.73
     Tycoon
    -0.65
     myself
    -0.65
     himself
    -0.63
     Guan
    -0.63
     Himself
    -0.61
    cussion
    -0.59
     Canaver
    -0.59
     extrad
    -0.59
    itialized
    -0.59
    POSITIVE LOGITS
     selves
    1.56
     counterparts
    1.43
     brethren
    0.95
    selves
    0.94
     histories
    0.90
     identities
    0.85
     careers
    0.83
     equivalents
    0.83
     tongues
    0.82
     asses
    0.82
    Activations Density 0.444%

    No Known Activations