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

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    © Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Feature Splitting for GPT2-Small
    3. GPT2-Small
    4. Residual Feature Splitting 12288
    5. 8-RES_FS12288-JB
    6. 9958
    Prev
    Next
    INDEX
    Explanations

    topics related to social or political issues, including nuclear proliferation, weight loss, due process, anonymity, freedom of religion, crime, segregation, advertising, religiosity, and religion

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs12288-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_12288
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    12,288
    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
    +0.30
    1.4%
    481
    +0.21
    1.0%
    454
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    288
    +0.30
    0.11
    481
    +0.21
    0.08
    439
    +0.10
    0.07
    Negative Logits
    Eva
    -0.54
    âĸij
    -0.54
    amina
    -0.54
    ammad
    -0.54
    ovie
    -0.52
    UFC
    -0.52
    than
    -0.51
    Ze
    -0.51
    ;;;;
    -0.51
    FTWARE
    -0.50
    POSITIVE LOGITS
     isn
    1.05
     tends
    1.05
     refers
    1.01
     exists
    0.99
     involves
    0.99
     occurs
    0.99
     is
    0.98
     dominates
    0.98
     consists
    0.97
     encompasses
    0.95
    Activations Density 0.681%

    No Known Activations