© 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. 12889
    Prev
    Next
    INDEX
    Explanations

    geographical features, specifically tectonic plates

    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
    289
    +0.14
    0.6%
    420
    +0.11
    0.5%
    179
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    289
    +0.14
    0.01
    76
    +0.11
    0.01
    575
    +0.11
    0.01
    Negative Logits
    vernment
    -1.03
    rians
    -0.84
    emouth
    -0.77
    etheless
    -0.74
    estern
    -0.73
    Plug
    -0.73
    sbm
    -0.72
    ITNESS
    -0.71
    ruary
    -0.70
    @#&
    -0.68
    POSITIVE LOGITS
     plates
    0.98
     plate
    0.95
    aus
    0.93
    plate
    0.92
    ographer
    0.90
    meal
    0.81
    ozo
    0.81
    cloth
    0.81
    lets
    0.74
    idon
    0.74
    Activations Density 0.018%

    No Known Activations