© 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 49152
    9. 8-RES_FS49152-JB
    11. 37434
    Prev
    Next
    INDEX
    Explanations

    phrases indicating various contexts of "getting out" or "escaping" from situations

    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
    373
    +0.15
    0.7%
    21
    +0.11
    0.5%
    172
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    553
    +0.15
    0.01
    666
    +0.11
    0.01
    270
    +0.09
    0.01
    Negative Logits
    pmwiki
    -0.76
    vol
    -0.62
    volume
    -0.61
    gee
    -0.60
    sports
    -0.59
    avery
    -0.59
    lish
    -0.57
    umbnails
    -0.57
    ocity
    -0.57
    ugg
    -0.56
    POSITIVE LOGITS
    reaching
    0.71
    ASY
    0.68
     towards
    0.63
    -------------
    0.63
     toward
    0.62
    adan
    0.62
    landers
    0.62
     salvation
    0.62
    isphere
    0.61
     homophobia
    0.60
    Activations Density 0.037%

    No Known Activations