© 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. 22766
    Prev
    Next
    INDEX
    Explanations

    phrases indicating instructions or follow-up actions

    oai_token-act-pair · gpt-4o-miniTriggered 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
    211
    +0.12
    0.6%
    218
    +0.10
    0.5%
    373
    +0.09
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    211
    +0.12
    0.01
    370
    +0.10
    0.01
    604
    +0.09
    0.01
    Negative Logits
     merch
    -0.58
    ufact
    -0.56
     robbers
    -0.55
    quartered
    -0.54
    osponsors
    -0.54
     beginnings
    -0.52
     intrins
    -0.51
    Footnote
    -0.50
     gearing
    -0.50
     liquids
    -0.50
    POSITIVE LOGITS
    thia
    0.77
    icol
    0.71
    OSE
    0.68
    andel
    0.66
    asse
    0.63
    skip
    0.62
    Finish
    0.62
    Cola
    0.61
     Cancel
    0.61
    Sorry
    0.60
    Activations Density 0.017%

    No Known Activations