© 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. 2921
    Prev
    Next
    INDEX
    Explanations

    advisory comments or recommendations to guide decision-making

    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
    640
    +0.17
    0.8%
    497
    +0.13
    0.6%
    440
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    6
    +0.17
    0.09
    640
    +0.13
    0.07
    440
    +0.11
    0.06
    Negative Logits
     mutation
    -0.74
     skeletal
    -0.74
     unaccount
    -0.73
     yet
    -0.68
     royalty
    -0.65
     inference
    -0.65
    awed
    -0.65
     boarded
    -0.65
     ancestry
    -0.64
     settled
    -0.64
    POSITIVE LOGITS
     Otherwise
    1.51
     Ideally
    1.37
     Doing
    1.33
     Then
    1.32
     Include
    1.26
     Especially
    1.25
     Alternatively
    1.24
     If
    1.22
     Use
    1.19
     Reduce
    1.19
    Activations Density 0.286%

    No Known Activations