© 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. 24459
    Prev
    Next
    INDEX
    Explanations

    commands and packages related to software management in a Linux environment

    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
    +0.14
    0.7%
    503
    +0.12
    0.6%
    591
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    284
    +0.14
    0.02
    503
    +0.12
    0.01
    373
    +0.11
    -0.01
    Negative Logits
    HUD
    -0.75
    vernment
    -0.73
     Pigs
    -0.70
    iov
    -0.65
    how
    -0.65
     Breath
    -0.65
    urrection
    -0.64
     mascara
    -0.63
    TED
    -0.63
    heit
    -0.63
    POSITIVE LOGITS
     unse
    0.66
     upstream
    0.63
     secure
    0.60
     linux
    0.60
     sacrific
    0.60
    master
    0.58
     root
    0.58
     gcc
    0.58
    author
    0.56
     authorizing
    0.56
    Activations Density 0.054%

    No Known Activations