© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    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. 24268
    Prev
    Next
    INDEX
    Explanations

    variations of the word "father."

    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
    314
    +0.17
    0.8%
    373
    +0.11
    0.5%
    356
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    288
    +0.17
    0.01
    314
    +0.11
    0.00
    48
    +0.11
    0.01
    Negative Logits
    ELS
    -0.70
    ured
    -0.70
     chef
    -0.63
     deaf
    -0.62
     pomp
    -0.58
    WS
    -0.57
    urious
    -0.55
     diabetic
    -0.55
     spite
    -0.55
     fries
    -0.55
    POSITIVE LOGITS
    apy
    1.60
    apist
    1.14
    nian
    1.13
    sburg
    1.12
    swer
    1.09
    s
    1.08
    ding
    1.05
    idge
    1.01
    mal
    0.95
    theless
    0.94
    Activations Density 0.019%

    No Known Activations