© 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. 24418
    Prev
    Next
    INDEX
    Explanations

    phrases and words related to political discourse and community concerns

    oai_token-act-pair · gpt-4o-miniTriggered 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
    640
    +0.23
    1.1%
    94
    +0.10
    0.5%
    244
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    640
    +0.23
    0.13
    94
    +0.10
    0.11
    539
    +0.09
    0.07
    Negative Logits
    utical
    -0.63
     Reincarnated
    -0.63
    uno
    -0.59
    doors
    -0.56
    sheets
    -0.56
    aeda
    -0.55
     Cig
    -0.55
    boro
    -0.54
     geography
    -0.54
     obscurity
    -0.53
    POSITIVE LOGITS
    Specifically
    0.90
    His
    0.86
    He
    0.84
    Writing
    0.80
    Speaking
    0.80
    Asked
    0.79
    "(
    0.78
    Scroll
    0.78
    Refer
    0.77
    PHOTOS
    0.76
    Activations Density 0.357%

    No Known Activations