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

    Neuronpedia

    APISteerSAE EvalsBlog/PodcastNEWSlackPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlog/PodcastGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Feature Splitting for GPT2-Small
    3. GPT2-Small
    4. Residual Feature Splitting 768
    5. 8-RES_FS768-JB
    6. 188
    Prev
    Next
    INDEX
    Explanations

    proper nouns and specific names of people, locations, or titles

    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
    274
    +0.15
    0.7%
    662
    +0.12
    0.6%
    393
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    274
    +0.15
    0.14
    662
    +0.12
    0.13
    598
    +0.10
    0.09
    Negative Logits
    emis
    -0.82
    cycles
    -0.72
    asures
    -0.71
    izont
    -0.68
    ciples
    -0.68
    boards
    -0.67
    hips
    -0.67
    ocks
    -0.66
    pots
    -0.66
    perse
    -0.65
    POSITIVE LOGITS
     pesky
    1.15
     fateful
    1.01
     same
    1.00
    cher
    0.99
     kind
    0.94
     translates
    0.87
     sort
    0.86
     mattered
    0.83
     particular
    0.83
     elusive
    0.82
    Activations Density 1.614%

    No Known Activations