© 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
    2. Joseph Bloom · Feature Splitting for GPT2-Small
    3. GPT2-Small
    4. Residual Feature Splitting 12288
    5. 8-RES_FS12288-JB
    6. 10719
    Prev
    Next
    INDEX
    Explanations

    discussions related to social issues, activism, and politics

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs12288-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_12288
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    12,288
    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
    721
    +0.09
    0.4%
    318
    +0.09
    0.4%
    715
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    43
    +0.09
    0.04
    380
    +0.09
    0.03
    429
    +0.09
    0.04
    Negative Logits
    OTH
    -0.68
    OTHER
    -0.66
    TeX
    -0.63
    usterity
    -0.63
    OND
    -0.63
    MpServer
    -0.61
    AMS
    -0.61
     (%
    -0.61
    RM
    -0.60
    Eye
    -0.60
    POSITIVE LOGITS
     however
    0.82
     huh
    0.74
     she
    0.74
     he
    0.73
     meanwhile
    0.71
     Julie
    0.71
     Jackie
    0.69
     moreover
    0.68
     Stra
    0.68
    dit
    0.67
    Activations Density 0.283%

    No Known Activations