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

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsinterp-engineNEWAPI 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. 14072
    Prev
    Next
    INDEX
    Explanations

    political terms and information related to U.S. Presidents, particularly Donald Trump and Barack Obama

    oai_token-act-pair · gpt-3.5-turboTriggered 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
    481
    +0.11
    0.5%
    168
    +0.10
    0.4%
    764
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    531
    +0.11
    0.02
    488
    +0.10
    0.02
    168
    +0.10
    0.02
    Negative Logits
    alog
    -0.81
    ostic
    -0.77
    MQ
    -0.71
    Ku
    -0.69
    isoft
    -0.65
    ologist
    -0.65
    asca
    -0.64
    Submit
    -0.64
    Unity
    -0.64
    alore
    -0.63
    POSITIVE LOGITS
     pledging
    0.95
    clinton
    0.82
    holders
    0.81
     swearing
    0.74
    holder
    0.72
    long
    0.71
    manship
    0.71
     levers
    0.70
     hopeful
    0.70
    wreck
    0.69
    Activations Density 0.045%

    No Known Activations