© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    1. Home
    3. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 1536
    9. 8-RES_FS1536-JB
    11. 81
    Prev
    Next
    INDEX
    Explanations

    words related to rules, instructions, and advice

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs1536-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    1,536
    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
    669
    +0.10
    0.4%
    643
    +0.09
    0.4%
    500
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    643
    +0.10
    0.09
    669
    +0.09
    0.09
    221
    +0.09
    0.06
    Negative Logits
    strength
    -0.72
     restored
    -0.71
     Enhancement
    -0.69
    ciation
    -0.69
     stabilized
    -0.68
    toc
    -0.68
    kind
    -0.66
    orld
    -0.64
    issance
    -0.63
    sort
    -0.63
    POSITIVE LOGITS
     undue
    1.05
     wasting
    1.03
     waste
    1.02
     bother
    1.00
     unnecessary
    0.99
     distractions
    0.98
     fuss
    0.97
     overly
    0.97
     stray
    0.96
     unnecessarily
    0.92
    Activations Density 2.217%

    No Known Activations