© 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 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. 17676
    Prev
    Next
    INDEX
    Explanations

    medical terms and statistics related to anesthesia awareness in surgical patients

    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
    339
    +0.13
    0.6%
    645
    +0.10
    0.5%
    622
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    339
    +0.13
    0.03
    622
    +0.10
    0.02
    645
    +0.09
    0.03
    Negative Logits
    cipline
    -0.74
    abad
    -0.73
    aleigh
    -0.72
     Himself
    -0.70
    another
    -0.69
    rea
    -0.69
    hyde
    -0.69
    especially
    -0.69
    literally
    -0.68
    ocracy
    -0.67
    POSITIVE LOGITS
     generations
    1.40
     incarn
    1.22
     iterations
    1.22
     installments
    1.14
     editions
    1.12
     administrations
    1.10
     incarnation
    1.09
    ebin
    1.09
     eras
    1.08
     versions
    1.04
    Activations Density 0.036%

    No Known Activations