© 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. 4953
    Prev
    Next
    INDEX
    Explanations

    words related to actions or activities that are prohibited or illegal

    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
    496
    +0.11
    0.5%
    242
    +0.11
    0.5%
    447
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    125
    +0.11
    0.03
    628
    +0.11
    0.03
    197
    +0.11
    0.03
    Negative Logits
    ochond
    -0.77
    ortun
    -0.67
     frustrations
    -0.66
    Progress
    -0.66
    ustration
    -0.64
    issance
    -0.63
    itar
    -0.63
    alky
    -0.61
    ushes
    -0.60
    rious
    -0.60
    POSITIVE LOGITS
     punishable
    0.94
     prohibited
    0.88
     violation
    0.85
     unless
    0.79
     permissible
    0.77
     prohibits
    0.75
     forbids
    0.75
     federally
    0.75
     grounds
    0.74
     violations
    0.72
    Activations Density 0.102%

    No Known Activations