© 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. 15694
    Prev
    Next
    INDEX
    Explanations

    phrases containing the pronoun "they" in the context of opinions or actions

    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
    393
    +0.10
    0.5%
    551
    +0.10
    0.5%
    566
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    566
    +0.10
    0.03
    477
    +0.10
    0.03
    354
    +0.10
    0.03
    Negative Logits
     Eps
    -0.81
     Cumber
    -0.71
     Dome
    -0.69
     Conquest
    -0.66
     Cobb
    -0.64
     Vine
    -0.64
     Marriage
    -0.63
     Gran
    -0.63
     Vog
    -0.62
    Posts
    -0.62
    POSITIVE LOGITS
    've
    1.22
    'd
    1.17
    're
    1.13
    'll
    0.94
     encount
    0.91
    selves
    0.89
     smelled
    0.89
     tasted
    0.89
     appre
    0.87
    zbollah
    0.87
    Activations Density 0.087%

    No Known Activations