© 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 12288
    9. 8-RES_FS12288-JB
    11. 554
    Prev
    Next
    INDEX
    Explanations

    phrases suggesting political debate

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs12288-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_12288
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    12,288
    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
    172
    +0.11
    0.6%
    671
    +0.11
    0.5%
    136
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    172
    +0.11
    0.08
    671
    +0.11
    0.06
    8
    +0.10
    0.07
    Negative Logits
     [+
    -0.71
    Therefore
    -0.68
     Cosponsors
    -0.66
    ItemImage
    -0.65
     '[
    -0.65
     therefore
    -0.63
     ItemLevel
    -0.63
     compan
    -0.63
    sbm
    -0.62
    ascript
    -0.62
    POSITIVE LOGITS
     booze
    0.72
     headlines
    0.61
     reckoning
    0.60
    ueller
    0.60
     nab
    0.60
     surprises
    0.57
     woes
    0.57
     downright
    0.57
     roses
    0.57
     celebrations
    0.56
    Activations Density 1.921%

    No Known Activations