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. 399
    Prev
    Next
    INDEX
    Explanations

    mentions of committees and investigations

    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
    644
    +0.12
    0.5%
    31
    +0.11
    0.5%
    681
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    31
    +0.12
    0.07
    146
    +0.11
    0.06
    247
    +0.10
    0.05
    Negative Logits
    XP
    -0.65
     Tide
    -0.63
    icio
    -0.63
    imm
    -0.62
     prest
    -0.62
     dx
    -0.61
     Univers
    -0.61
    Zip
    -0.61
    paralle
    -0.61
    ulia
    -0.60
    POSITIVE LOGITS
     chaired
    1.25
     hearings
    1.14
    ittee
    1.14
     chairman
    1.09
     convened
    1.09
     chair
    1.03
     tasked
    0.97
     Chair
    0.95
     investigating
    0.94
     Chairman
    0.94
    Activations Density 0.613%

    No Known Activations