© 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. 14478
    Prev
    Next
    INDEX
    Explanations

    references to bills and legislative actions

    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%
    481
    +0.11
    0.5%
    72
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    524
    +0.12
    0.05
    288
    +0.11
    0.06
    575
    +0.10
    0.05
    Negative Logits
    othe
    -0.85
    yss
    -0.81
    姫
    -0.75
    pmwiki
    -0.75
    Jr
    -0.71
    alle
    -0.68
    ixel
    -0.67
    ocytes
    -0.66
     keyboards
    -0.65
    asin
    -0.65
    POSITIVE LOGITS
     mandate
    0.80
     mandated
    0.77
     overseen
    0.76
     issued
    0.74
     backlog
    0.71
     approved
    0.71
    undown
    0.70
    naire
    0.70
     roundup
    0.69
     jointly
    0.69
    Activations Density 0.225%

    No Known Activations