© 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. 10039
    Prev
    Next
    INDEX
    Explanations

    notable items or topics within a larger context

    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
    151
    +0.17
    0.8%
    540
    +0.11
    0.5%
    284
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    151
    +0.17
    0.03
    429
    +0.11
    0.02
    101
    +0.11
    0.02
    Negative Logits
    hell
    -0.89
    prep
    -0.83
    claimer
    -0.80
    yrinth
    -0.74
    cock
    -0.73
    amar
    -0.72
    alyst
    -0.72
    ©¶æ
    -0.71
    nan
    -0.69
    lyn
    -0.69
    POSITIVE LOGITS
     exception
    1.09
     exceptions
    1.06
     absent
    1.01
     landmarks
    0.95
     absence
    0.95
     accomplishments
    0.92
     differences
    0.88
     successes
    0.86
     contributions
    0.83
     accomplishment
    0.83
    Activations Density 0.050%

    No Known Activations