© 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 24576
    9. 8-RES_FS24576-JB
    11. 17811
    Prev
    Next
    INDEX
    Explanations

    words related to deviation or divergence from expected norms or directions

    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
    288
    +0.12
    0.6%
    360
    +0.09
    0.4%
    286
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    4
    +0.12
    0.01
    211
    +0.09
    0.01
    286
    +0.09
    0.01
    Negative Logits
    xual
    -0.83
     Twain
    -0.77
     Rouge
    -0.76
    terday
    -0.73
     Moonlight
    -0.73
     Tempest
    -0.66
     Beir
    -0.66
     Objective
    -0.65
    tone
    -0.65
    ¥µ
    -0.63
    POSITIVE LOGITS
    olved
    1.45
    irtual
    1.41
    olution
    1.36
    olve
    1.30
    iating
    1.28
    iates
    1.27
    iant
    1.26
    iated
    1.25
    iants
    1.22
    ices
    1.21
    Activations Density 0.008%

    No Known Activations