© 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. 5204
    Prev
    Next
    INDEX
    Explanations

    questions inviting analysis or discussion

    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
    512
    +0.12
    0.5%
    314
    +0.12
    0.5%
    566
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    566
    +0.12
    0.04
    512
    +0.12
    0.04
    344
    +0.11
    0.04
    Negative Logits
    etting
    -0.69
     wounding
    -0.60
    typ
    -0.59
    tip
    -0.58
    rift
    -0.58
    being
    -0.56
     unfolding
    -0.55
    etter
    -0.55
    lig
    -0.54
     pledging
    -0.54
    POSITIVE LOGITS
     recourse
    1.06
     been
    0.98
     gotten
    0.92
     conversations
    0.88
     undergone
    0.87
     gone
    0.84
    gotten
    0.83
     access
    0.81
    stood
    0.81
    azeera
    0.81
    Activations Density 0.106%

    No Known Activations