© 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 EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 49152
    9. 8-RES_FS49152-JB
    11. 43126
    Prev
    Next
    INDEX
    Explanations

    repeated instances of specific suffixes in words

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs49152-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    49,152
    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
    442
    +0.12
    0.5%
    373
    +0.11
    0.5%
    659
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    730
    +0.12
    0.01
    373
    +0.11
    -0.00
    284
    +0.10
    0.01
    Negative Logits
    ctica
    -1.00
    aneously
    -0.74
     arteries
    -0.74
    tions
    -0.63
     Carnage
    -0.61
     Saddam
    -0.61
     Tsukuyomi
    -0.61
    ctors
    -0.60
     Salvation
    -0.59
     taxpayers
    -0.58
    POSITIVE LOGITS
    inki
    1.18
    kamp
    1.05
    sein
    0.99
    chel
    0.98
    inger
    0.96
    sung
    0.93
    mann
    0.92
    mith
    0.88
    linger
    0.88
    dor
    0.88
    Activations Density 0.012%

    No Known Activations