© 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. 5519
    Prev
    Next
    INDEX
    Explanations

    mentions of information or items put on display or showcased

    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
    481
    +0.18
    0.8%
    762
    +0.11
    0.5%
    430
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    762
    +0.18
    0.04
    288
    +0.11
    0.03
    575
    +0.10
    0.03
    Negative Logits
    ARDIS
    -0.67
    ritical
    -0.66
    anche
    -0.66
    yssey
    -0.65
     Patriarch
    -0.64
    lez
    -0.64
    ensus
    -0.63
    laus
    -0.63
    pell
    -0.62
    auts
    -0.61
    POSITIVE LOGITS
     tonight
    0.78
     today
    0.74
     TODAY
    0.67
     NOW
    0.65
     Tonight
    0.65
     shelves
    0.65
     tomorrow
    0.64
    Ø©
    0.63
     HERE
    0.63
    Condition
    0.60
    Activations Density 0.101%

    No Known Activations