© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI 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. 20319
    Prev
    Next
    INDEX
    Explanations

    statements of contrast or contradiction

    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
    452
    +0.10
    0.4%
    22
    +0.09
    0.4%
    370
    +0.08
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    638
    +0.10
    0.04
    431
    +0.09
    0.03
    713
    +0.08
    0.03
    Negative Logits
    +.
    -0.58
     concess
    -0.58
     frequ
    -0.57
     hemor
    -0.57
     cess
    -0.56
     futile
    -0.55
     vain
    -0.54
     classics
    -0.54
     Moroc
    -0.54
     heels
    -0.53
    POSITIVE LOGITS
     illustrates
    0.83
     differs
    0.83
     varies
    0.82
     is
    0.81
     depends
    0.81
     boils
    0.76
     involves
    0.75
    SPONSORED
    0.74
     underscores
    0.68
     differed
    0.68
    Activations Density 0.133%

    No Known Activations