© 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. 14367
    Prev
    Next
    INDEX
    Explanations

    religious or political causes

    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
    59
    +0.14
    0.6%
    620
    +0.10
    0.4%
    127
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    59
    +0.14
    0.02
    455
    +0.10
    0.02
    255
    +0.10
    0.02
    Negative Logits
    Ku
    -0.76
     Seasons
    -0.68
    atari
    -0.66
     Storm
    -0.65
    PDATE
    -0.65
    egu
    -0.65
     Leopard
    -0.64
    eatures
    -0.63
     Penet
    -0.63
     Sachs
    -0.62
    POSITIVE LOGITS
     cele
    1.47
     celeb
    0.81
    umed
    0.79
    way
    0.79
    ality
    0.74
    wagon
    0.72
    usa
    0.72
    cause
    0.72
    llor
    0.71
    unity
    0.71
    Activations Density 0.016%

    No Known Activations