© 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 49152
    9. 8-RES_FS49152-JB
    11. 41568
    Prev
    Next
    INDEX
    Explanations

    mentions of Russia and its government, as well as references to Russian individuals and interference

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs49152-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    49,152
    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
    682
    +0.11
    0.5%
    668
    +0.10
    0.5%
    621
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    67
    +0.11
    0.03
    682
    +0.10
    0.03
    621
    +0.10
    0.03
    Negative Logits
    dash
    -0.78
    erick
    -0.77
    Thom
    -0.77
    draw
    -0.75
     Dublin
    -0.75
     Dickinson
    -0.73
     Towns
    -0.72
     Kier
    -0.71
    enment
    -0.71
    liam
    -0.71
    POSITIVE LOGITS
     collusion
    1.11
     meddling
    1.10
     hacking
    1.02
     intelligence
    0.99
     hoax
    0.96
     disinformation
    0.96
     espionage
    0.94
     dossier
    0.93
     interference
    0.92
    intelligence
    0.91
    Activations Density 0.037%

    No Known Activations