© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 24576
    9. 8-RES_FS24576-JB
    11. 18295
    Prev
    Next
    INDEX
    Explanations

    statements emphasizing strong opinions or reactions on political or social issues

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs24576-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    24,576
    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
    561
    +0.11
    0.5%
    497
    +0.10
    0.5%
    512
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    8
    +0.11
    0.05
    512
    +0.10
    0.04
    49
    +0.10
    0.04
    Negative Logits
    "},"
    -0.73
    Scroll
    -0.68
     rush
    -0.63
     Afterwards
    -0.62
    Alert
    -0.61
     whilst
    -0.61
    inav
    -0.60
    Cooldown
    -0.58
     denies
    -0.56
    Is
    -0.56
    POSITIVE LOGITS
     merely
    0.87
     simply
    0.76
    Ħ¢
    0.70
    ngth
    0.69
    olitical
    0.67
    staking
    0.66
    ãĥł
    0.66
    archy
    0.63
    INESS
    0.63
    quartered
    0.63
    Activations Density 0.204%

    No Known Activations