© 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. 5177
    Prev
    Next
    INDEX
    Explanations

    phrases related to political elections and governance

    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
    +0.24
    1.1%
    719
    +0.12
    0.5%
    181
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    575
    +0.24
    0.09
    762
    +0.12
    0.09
    288
    +0.09
    0.09
    Negative Logits
     cameo
    -0.66
     Boat
    -0.64
    arij
    -0.63
     charm
    -0.60
    jah
    -0.58
    ¶ħ
    -0.58
     shoulders
    -0.57
     oy
    -0.57
    wered
    -0.57
     Hawks
    -0.56
    POSITIVE LOGITS
     safely
    1.06
     abroad
    1.06
     efficiently
    1.03
     securely
    0.90
     withd
    0.88
     outdoors
    0.87
     cheaply
    0.87
     without
    0.86
     anonymously
    0.85
     electronically
    0.85
    Activations Density 0.551%

    No Known Activations