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 24576
    5. 8-RES_FS24576-JB
    6. 23377
    Prev
    Next
    INDEX
    Explanations

    phrases involving the conjunction "and"

    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
    203
    +0.13
    0.6%
    282
    +0.10
    0.4%
    481
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    203
    +0.13
    0.05
    128
    +0.10
    0.04
    713
    +0.09
    0.04
    Negative Logits
    Times
    -0.69
    emi
    -0.64
    "],
    -0.63
    psc
    -0.60
    otto
    -0.60
    iak
    -0.59
    iatus
    -0.58
     {:
    -0.57
    spot
    -0.56
     Strikes
    -0.56
    POSITIVE LOGITS
     consequently
    1.31
     preferably
    1.25
     furthermore
    1.25
     therefore
    1.23
     indeed
    1.19
     optionally
    1.18
     hence
    1.15
     moreover
    1.14
     especially
    1.14
     possibly
    1.13
    Activations Density 0.194%

    No Known Activations