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

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    1. Home
    3. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 768
    9. 8-RES_FS768-JB
    11. 714
    Prev
    Next
    INDEX
    Explanations

    words related to physical attributes such as weight, exercise, appearance, and other related concepts

    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
    447
    +0.27
    1.3%
    481
    +0.13
    0.6%
    762
    +0.12
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    575
    +0.27
    0.29
    288
    +0.13
    0.31
    762
    +0.12
    0.27
    Negative Logits
     specificity
    -0.60
    audi
    -0.57
     collaboration
    -0.51
     pestic
    -0.50
    ãĥª
    -0.49
    stood
    -0.49
     coincidence
    -0.49
     convergence
    -0.48
     birth
    -0.48
     narrowed
    -0.47
    POSITIVE LOGITS
     whilst
    1.16
     while
    1.09
     whenever
    1.06
     WITHOUT
    1.05
     during
    1.03
     without
    1.02
     because
    0.97
     abroad
    0.97
     wherever
    0.97
     before
    0.95
    Activations Density 7.744%

    No Known Activations