© 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. 18022
    Prev
    Next
    INDEX
    Explanations

    action-related events in a narrative, particularly those involving witnessing or hearing about something happening to others

    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
    447
    +0.12
    0.6%
    218
    +0.10
    0.5%
    765
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    403
    +0.12
    0.04
    486
    +0.10
    0.03
    765
    +0.10
    0.03
    Negative Logits
    xus
    -0.71
    bia
    -0.70
    ongyang
    -0.68
    trap
    -0.67
    vic
    -0.66
    buster
    -0.66
    nesia
    -0.65
    mind
    -0.65
    esan
    -0.64
    sis
    -0.64
    POSITIVE LOGITS
     something
    0.86
     flashes
    0.83
     nothing
    0.82
     someone
    0.82
     faint
    0.81
     him
    0.78
     firsthand
    0.78
     footprints
    0.78
     silhou
    0.76
     similarities
    0.75
    Activations Density 0.087%

    No Known Activations