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. 305
    Prev
    Next
    INDEX
    Explanations

    steps or instructions

    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
    102
    +0.21
    1.1%
    566
    +0.18
    0.9%
    55
    +0.16
    0.8%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    102
    +0.21
    0.31
    55
    +0.18
    0.30
    566
    +0.16
    0.27
    Negative Logits
     occurs
    -0.81
    osate
    -0.78
    inav
    -0.77
    ulence
    -0.74
     Highlights
    -0.72
     lasts
    -0.71
     legality
    -0.70
    ossom
    -0.70
     Ensure
    -0.70
     flourish
    -0.69
    POSITIVE LOGITS
     aware
    1.37
     afraid
    1.34
     able
    1.32
     willing
    1.27
     unable
    1.25
     thankful
    1.21
     glad
    1.20
     unaware
    1.16
     obligated
    1.16
     fortunate
    1.16
    Activations Density 3.432%

    No Known Activations