© 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. 15767
    Prev
    Next
    INDEX
    Explanations

    actions or events involving confrontation and physical altercations

    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
    314
    +0.19
    0.9%
    183
    +0.13
    0.6%
    566
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    183
    +0.19
    0.06
    372
    +0.13
    0.05
    282
    +0.11
    0.05
    Negative Logits
     Archdemon
    -0.64
     depiction
    -0.59
     inhabit
    -0.58
    velt
    -0.58
     endeavour
    -0.57
    iege
    -0.56
     endeavors
    -0.53
     Alic
    -0.53
     Madness
    -0.53
    mare
    -0.52
    POSITIVE LOGITS
     rid
    1.33
    cloneembedreportprint
    0.95
     lucky
    0.93
    cha
    0.88
     acquainted
    0.88
     caught
    0.82
     bitten
    0.82
    DragonMagazine
    0.81
     distracted
    0.81
     stuck
    0.81
    Activations Density 0.105%

    No Known Activations