© 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 12288
    9. 8-RES_FS12288-JB
    11. 10686
    Prev
    Next
    INDEX
    Explanations

    references to buses or public transportation systems

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs12288-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_12288
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    12,288
    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
    484
    +0.12
    0.6%
    417
    +0.11
    0.5%
    288
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    484
    +0.12
    0.03
    343
    +0.11
    0.03
    115
    +0.11
    0.03
    Negative Logits
     marrow
    -0.72
     Centauri
    -0.71
     Martial
    -0.66
    theless
    -0.65
     Accuracy
    -0.64
     reef
    -0.63
    aneous
    -0.61
     limb
    -0.61
    AFTA
    -0.59
    arily
    -0.58
    POSITIVE LOGITS
    loads
    1.29
    INESS
    1.14
    load
    1.12
    hel
    1.04
    boys
    0.98
    kers
    0.93
    hy
    0.91
    engers
    0.88
    driver
    0.88
    ker
    0.85
    Activations Density 0.081%

    No Known Activations