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. 302
    Prev
    Next
    INDEX
    Explanations

    details related to violent crimes and legal proceedings

    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
    386
    +0.14
    0.6%
    314
    +0.12
    0.5%
    628
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    386
    +0.14
    0.14
    628
    +0.12
    0.10
    86
    +0.11
    0.10
    Negative Logits
    NetMessage
    -0.79
    layout
    -0.75
     Remastered
    -0.75
    echo
    -0.74
    ysc
    -0.72
    optim
    -0.72
    phies
    -0.71
    soDeliveryDate
    -0.71
     optim
    -0.70
    clips
    -0.70
    POSITIVE LOGITS
     felony
    1.60
     misdemeanor
    1.52
     misdem
    1.47
     criminally
    1.46
     manslaughter
    1.46
     offences
    1.39
     perjury
    1.38
     punishable
    1.38
     convicted
    1.38
     criminal
    1.34
    Activations Density 4.583%

    No Known Activations