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

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    1. Home
    3. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 768
    9. 8-RES_FS768-JB
    11. 710
    Prev
    Next
    INDEX
    Explanations

    words related to negative political opinions and controversy

    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
    709
    +0.12
    0.6%
    621
    +0.11
    0.5%
    130
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    621
    +0.12
    0.10
    469
    +0.11
    0.07
    402
    +0.10
    0.08
    Negative Logits
    utenant
    -0.72
     1915
    -0.69
     1962
    -0.69
     2024
    -0.65
    ony
    -0.65
    Url
    -0.62
     ç¥ŀ
    -0.62
    opus
    -0.61
     Retrieved
    -0.61
    iste
    -0.61
    POSITIVE LOGITS
    worldly
    2.10
     aspects
    0.99
    wise
    0.98
     avenues
    0.96
     factors
    0.96
     types
    0.93
     kinds
    0.92
     countries
    0.92
     disciplines
    0.89
    world
    0.87
    Activations Density 2.679%

    No Known Activations