© 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. 10193
    Prev
    Next
    INDEX
    Explanations

    sentences related to crimes or criminal investigations

    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
    102
    +0.10
    0.5%
    384
    +0.09
    0.4%
    718
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    102
    +0.10
    0.05
    566
    +0.09
    0.05
    55
    +0.09
    0.05
    Negative Logits
     VIDEOS
    -0.75
    Guide
    -0.67
    alogy
    -0.66
    izes
    -0.66
    rones
    -0.64
    tymology
    -0.63
    ively
    -0.63
    Today
    -0.62
    Tokens
    -0.61
     Brach
    -0.61
    POSITIVE LOGITS
     previously
    1.06
    foreseen
    0.86
     booked
    0.81
    cffffcc
    0.81
     briefed
    0.81
     ordered
    0.80
     instructed
    0.80
     exha
    0.80
     dormant
    0.80
     notified
    0.79
    Activations Density 0.110%

    No Known Activations