© 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. 3177
    Prev
    Next
    INDEX
    Explanations

    information related to tragic events or disasters, specifically focusing on incidents and occurrences

    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
    229
    +0.13
    0.6%
    637
    +0.12
    0.5%
    537
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    637
    +0.13
    0.05
    362
    +0.12
    0.06
    121
    +0.11
    0.05
    Negative Logits
     maxim
    -0.72
    galitarian
    -0.70
    dit
    -0.69
    hair
    -0.68
    ledge
    -0.66
    CHO
    -0.65
    Opt
    -0.65
    */(
    -0.65
    ologue
    -0.64
     respectfully
    -0.64
    POSITIVE LOGITS
     southwestern
    0.83
     perpetrated
    0.83
     suspected
    0.82
     southeastern
    0.82
     aftermath
    0.80
     spate
    0.80
     incidents
    0.80
     spree
    0.80
     unsolved
    0.79
    etimes
    0.79
    Activations Density 0.414%

    No Known Activations