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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & 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. 522
    Prev
    Next
    INDEX
    Explanations

    information on government policies and initiatives, as well as discussions on historical and cultural practices around the world

    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
    6
    +0.12
    0.6%
    667
    +0.10
    0.4%
    641
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    6
    +0.12
    0.19
    107
    +0.10
    0.15
    308
    +0.10
    0.12
    Negative Logits
    conservancy
    -0.75
     decomp
    -0.73
     hosting
    -0.73
    weights
    -0.73
     downs
    -0.72
     descended
    -0.70
     favour
    -0.70
     envelop
    -0.69
    gone
    -0.69
     untouched
    -0.69
    POSITIVE LOGITS
    ONSORED
    1.16
    However
    1.07
     Unfortunately
    1.04
     Also
    1.01
    Therefore
    1.00
    Also
    1.00
     However
    0.99
     Additionally
    0.99
     Similarly
    0.97
     Furthermore
    0.97
    Activations Density 2.312%

    No Known Activations