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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsBlogSlackPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Feature Splitting for GPT2-Small
    3. GPT2-Small
    4. Residual Feature Splitting 49152
    5. 8-RES_FS49152-JB
    6. 12193
    Prev
    Next
    INDEX
    Explanations

    instances of the word "delight" and variations of "indulge."

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs49152-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    49,152
    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
    288
    +0.13
    0.6%
    151
    +0.12
    0.5%
    465
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    151
    +0.13
    0.01
    271
    +0.12
    0.01
    317
    +0.10
    0.01
    Negative Logits
    ework
    -0.75
    papers
    -0.71
    agnetic
    -0.69
    DragonMagazine
    -0.68
     Patriarch
    -0.64
     Skydragon
    -0.64
     ($)
    -0.63
    assian
    -0.63
    kees
    -0.62
    antam
    -0.61
    POSITIVE LOGITS
    iously
    0.97
    fully
    0.97
    ãĥĦ
    0.88
    ously
    0.84
     delight
    0.79
     indul
    0.78
    rences
    0.74
     pleasures
    0.72
    MQ
    0.72
    UGE
    0.71
    Activations Density 0.010%

    No Known Activations