Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Feature Splitting for GPT2-Small
    3. GPT2-Small
    4. Residual Feature Splitting 24576
    5. 8-RES_FS24576-JB
    6. 1011
    Prev
    Next
    INDEX
    Explanations

    phrases related to political issues and controversial topics

    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
    401
    +0.12
    0.5%
    289
    +0.11
    0.5%
    129
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    763
    +0.12
    0.03
    397
    +0.11
    0.03
    737
    +0.10
    0.02
    Negative Logits
    ridor
    -0.72
    emale
    -0.68
    sha
    -0.65
    aza
    -0.65
    anmar
    -0.63
    ser
    -0.63
    eda
    -0.62
    sted
    -0.61
    ļéĨĴ
    -0.61
    undown
    -0.61
    POSITIVE LOGITS
    pires
    0.90
     whatsoever
    0.74
     chooses
    0.73
    aign
    0.71
     decides
    0.71
     circumstances
    0.69
     misc
    0.68
    ifice
    0.68
     deems
    0.67
     faults
    0.65
    Activations Density 0.120%

    No Known Activations