© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it
    Neuronpedia
    Jacobian LensNEW
    Natural Language
    Autoencoders
    MAY 26    Assistant AxisJAN 26Circuit TracerAPR 26SteerSAE EvalsExportsinterp-engineNEW 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. 14861
    Prev
    Next
    INDEX
    Explanations

    warnings or potential risks in the text

    oai_token-act-pair · gpt-3.5-turbo
    Triggered 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
    Neuron Alignment
    Index
    Value
    % of L₁
    373
    +0.16
    0.7%
    76
    +0.11
    0.5%
    334
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    153
    +0.16
    0.04
    436
    +0.11
    0.03
    637
    +0.10
    0.03
    Negative Logits
    ratulations
    -0.71
    keepers
    -0.69
    keeper
    -0.68
    phabet
    -0.67
    avorite
    -0.60
     congratulations
    -0.60
     Invaders
    -0.60
    renheit
    -0.59
    inis
    -0.58
    gre
    -0.58
    POSITIVE LOGITS
     behest
    1.23
     expense
    1.11
     intervals
    1.03
     intersections
    0.98
    point
    0.98
     point
    0.96
     outset
    0.95
     discretion
    0.94
     glance
    0.94
     moment
    0.94
    Activations Density 0.070%

    No Known Activations