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 3072
    5. 8-RES_FS3072-JB
    6. 2121
    Prev
    Next
    INDEX
    Explanations

    phrases related to physical barriers or obstacles that impede movement or progress

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs3072-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    3,072
    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
    3
    +0.10
    0.4%
    489
    +0.09
    0.4%
    632
    +0.08
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    298
    +0.10
    0.07
    362
    +0.09
    0.08
    608
    +0.08
    0.09
    Negative Logits
    olate
    -0.82
    olar
    -0.67
    arse
    -0.65
    oy
    -0.64
    lat
    -0.63
    sonian
    -0.63
    iquette
    -0.62
    elman
    -0.61
    worn
    -0.61
    char
    -0.61
    POSITIVE LOGITS
     access
    1.03
     accessing
    1.00
     progress
    0.99
     entry
    0.96
     growth
    0.88
     uptake
    0.87
     indefinitely
    0.87
     Entry
    0.86
     deportation
    0.84
     flow
    0.84
    Activations Density 1.608%

    No Known Activations