© 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 EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 49152
    9. 8-RES_FS49152-JB
    11. 43012
    Prev
    Next
    INDEX
    Explanations

    instances of the word "write" and its variations, indicating a focus on writing-related actions

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs49152-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    49,152
    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
    644
    +0.11
    0.5%
    167
    +0.09
    0.4%
    461
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    644
    +0.11
    0.03
    725
    +0.09
    0.03
    85
    +0.09
    0.03
    Negative Logits
    Ĭ±
    -0.94
    agara
    -0.76
    arov
    -0.65
    rolet
    -0.65
     Ukrain
    -0.64
    hon
    -0.64
    EGA
    -0.64
     Unsure
    -0.63
    atched
    -0.63
    RELATED
    -0.61
    POSITIVE LOGITS
     poems
    0.95
     poetry
    0.91
    itatively
    0.89
    writing
    0.87
    acters
    0.87
     essays
    0.86
     screenplay
    0.84
     letters
    0.81
     lyrics
    0.80
     aloud
    0.78
    Activations Density 0.037%

    No Known Activations