Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsBlogSlackPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Feature Splitting for GPT2-Small
    3. GPT2-Small
    4. Residual Feature Splitting 768
    5. 8-RES_FS768-JB
    6. 346
    Prev
    Next
    INDEX
    Explanations

    words related to technology products and features

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs768-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    768
    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
    459
    +0.12
    0.5%
    12
    +0.11
    0.5%
    194
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    665
    +0.12
    0.16
    508
    +0.11
    0.11
    12
    +0.10
    0.09
    Negative Logits
     Deaths
    -0.79
     capit
    -0.73
     Rebell
    -0.73
    fell
    -0.72
     Reverend
    -0.71
    sted
    -0.70
     Patri
    -0.70
     Reck
    -0.69
     Revision
    -0.68
     underest
    -0.68
    POSITIVE LOGITS
     seamlessly
    1.38
     efficiently
    1.27
     customizable
    1.24
     customized
    1.18
     securely
    1.18
     effortlessly
    1.17
     seamless
    1.13
     interact
    1.12
     offline
    1.10
     multimedia
    1.07
    Activations Density 8.635%

    No Known Activations