© 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 49152
    9. 8-RES_FS49152-JB
    11. 45614
    Prev
    Next
    INDEX
    Explanations

    emotional reactions expressed through exclamations or rhetorical questions

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs49152-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    49,152
    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
    596
    +0.11
    0.5%
    393
    +0.11
    0.5%
    474
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    391
    +0.11
    0.02
    596
    +0.11
    0.02
    389
    +0.10
    0.02
    Negative Logits
    vale
    -0.90
    ourse
    -0.87
    etheless
    -0.83
     guiActiveUn
    -0.77
    isSpecialOrderable
    -0.74
    eatures
    -0.71
    staking
    -0.70
    interstitial
    -0.70
    DragonMagazine
    -0.69
    inction
    -0.68
    POSITIVE LOGITS
     please
    0.96
     oh
    0.91
     yeah
    0.91
     why
    0.90
     huh
    0.85
     hurry
    0.81
     maybe
    0.79
     let
    0.78
     WHY
    0.77
     WHAT
    0.74
    Activations Density 0.040%

    No Known Activations