© 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. 2697
    Prev
    Next
    INDEX
    Explanations

    mentions of organized efforts or initiatives with an intended purpose

    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
    288
    +0.16
    0.7%
    447
    +0.12
    0.5%
    723
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    247
    +0.16
    0.03
    723
    +0.12
    0.03
    644
    +0.10
    0.03
    Negative Logits
    orgetown
    -0.93
    ĪĴ
    -0.78
     decaying
    -0.70
    ãĥĥãĥī
    -0.66
     Flo
    -0.65
     lowly
    -0.62
    sembly
    -0.62
    ŃĶ
    -0.62
    essional
    -0.62
     Lumpur
    -0.62
    POSITIVE LOGITS
     against
    1.05
     waged
    0.97
    horse
    0.93
     campaigns
    0.90
     undertaken
    0.88
     slogan
    0.85
     manager
    0.81
    ulations
    0.80
    campaign
    0.79
    against
    0.79
    Activations Density 0.036%

    No Known Activations