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 768
    5. 8-RES_FS768-JB
    6. 587
    Prev
    Next
    INDEX
    Explanations

    superlatives and evaluations such as "most," "difficult," "favorite," "best," "worst," "consistent," "irritating," "popular," "powerful," and "prominent" in a text

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs768-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    768
    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
    194
    +0.12
    0.5%
    719
    +0.10
    0.4%
    339
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    194
    +0.12
    0.10
    190
    +0.10
    0.10
    451
    +0.09
    0.08
    Negative Logits
    skirts
    -0.78
    kas
    -0.71
    thur
    -0.71
    rompt
    -0.65
    pload
    -0.63
    heid
    -0.63
    selves
    -0.63
    blocks
    -0.63
    roth
    -0.63
    too
    -0.63
    POSITIVE LOGITS
     imaginable
    0.99
     conceivable
    0.93
     possible
    0.86
     practicable
    0.79
    seller
    0.79
     ever
    0.77
     Wanted
    0.75
    liest
    0.71
     notable
    0.71
    ream
    0.70
    Activations Density 2.122%

    No Known Activations