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

    Neuronpedia

    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. 8721
    Prev
    Next
    INDEX
    Explanations

    phrases related to criticism, controversy, or negative events

    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
    480
    +0.42
    2.1%
    481
    +0.23
    1.2%
    693
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    480
    +0.42
    0.01
    266
    +0.23
    0.08
    481
    +0.09
    0.09
    Negative Logits
    ortality
    -0.63
    cellaneous
    -0.58
    aceae
    -0.58
     oneself
    -0.57
     srfAttach
    -0.57
     Alternatively
    -0.57
    OTAL
    -0.56
     depended
    -0.54
     myself
    -0.54
     Died
    -0.54
    POSITIVE LOGITS
     bullying
    0.75
     crackdown
    0.74
     antics
    0.71
     mishand
    0.71
     allegations
    0.68
     transgender
    0.68
     encro
    0.68
     perceived
    0.68
     accusations
    0.68
     scandal
    0.67
    Activations Density 0.948%

    No Known Activations