© 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 24576
    9. 8-RES_FS24576-JB
    11. 11045
    Prev
    Next
    INDEX
    Explanations

    words related to carrying objects

    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
    290
    +0.11
    0.5%
    615
    +0.09
    0.4%
    85
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    290
    +0.11
    0.03
    34
    +0.09
    0.02
    85
    +0.09
    0.03
    Negative Logits
    ellen
    -0.77
     Skydragon
    -0.75
    reddits
    -0.74
    chrom
    -0.72
     Seym
    -0.70
    Pie
    -0.67
    ederation
    -0.67
    clair
    -0.65
    hemer
    -0.64
    esters
    -0.62
    POSITIVE LOGITS
     loads
    0.84
     forward
    0.84
    weight
    0.82
     weight
    0.81
    forward
    0.78
     carrier
    0.78
    over
    0.77
     carriers
    0.75
    weights
    0.75
    overs
    0.74
    Activations Density 0.027%

    No Known Activations