© 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. 14166
    Prev
    Next
    INDEX
    Explanations

    words related to politics and economics, particularly in a global context

    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
    659
    +0.12
    0.5%
    279
    +0.11
    0.5%
    205
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    659
    +0.12
    0.06
    205
    +0.11
    0.05
    284
    +0.10
    0.05
    Negative Logits
    ENTION
    -0.70
    untled
    -0.68
    advertisement
    -0.67
    YC
    -0.66
    SPONSORED
    -0.65
    PHOTOS
    -0.65
    Making
    -0.64
    ModLoader
    -0.64
    ymm
    -0.63
    ALK
    -0.62
    POSITIVE LOGITS
    ño
    1.19
    ñ
    1.06
    ppa
    0.98
    ña
    0.95
     ni
    0.91
    ÅŁ
    0.90
     tion
    0.87
    zza
    0.86
     il
    0.86
     pa
    0.84
    Activations Density 0.113%

    No Known Activations