© 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. 22711
    Prev
    Next
    INDEX
    Explanations

    the word "which" to identify relative clauses or additional information in sentences

    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
    373
    +0.16
    0.7%
    471
    +0.10
    0.5%
    572
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    757
    +0.16
    0.01
    471
    +0.10
    0.01
    236
    +0.09
    0.01
    Negative Logits
    politics
    -0.71
    baby
    -0.71
    marg
    -0.71
    fo
    -0.71
    apult
    -0.69
    ma
    -0.69
    vous
    -0.66
    viks
    -0.66
    ben
    -0.66
    bt
    -0.66
    POSITIVE LOGITS
    soever
    0.74
     dearly
    0.74
     there
    0.72
    earch
    0.69
     comparisons
    0.66
     we
    0.66
     assessments
    0.64
     considerable
    0.64
     inquiries
    0.64
     evaluations
    0.63
    Activations Density 0.019%

    No Known Activations