© 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. 15047
    Prev
    Next
    INDEX
    Explanations

    instances of the word "back" being used as a metaphorical reference to revisiting or considering something again

    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
    242
    +0.11
    0.5%
    674
    +0.09
    0.4%
    760
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    242
    +0.11
    0.03
    760
    +0.09
    0.02
    619
    +0.09
    0.02
    Negative Logits
    anooga
    -0.66
    orst
    -0.61
     Us
    -0.56
     Donation
    -0.55
     Crowd
    -0.54
    rowd
    -0.53
    Specifically
    -0.53
     notoriously
    -0.53
     ideal
    -0.52
    ck
    -0.52
    POSITIVE LOGITS
    packing
    1.07
    stab
    1.03
    dated
    0.96
    packs
    0.95
    gam
    0.92
    doors
    0.91
    tracking
    0.87
    door
    0.85
    wards
    0.85
    home
    0.84
    Activations Density 0.032%

    No Known Activations