© 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 49152
    9. 8-RES_FS49152-JB
    11. 43109
    Prev
    Next
    INDEX
    Explanations

    phrases indicating the purpose or reason for actions

    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
    369
    +0.15
    0.7%
    583
    +0.12
    0.5%
    152
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    583
    +0.15
    0.02
    287
    +0.12
    0.02
    369
    +0.11
    0.02
    Negative Logits
     gorge
    -0.79
     hooked
    -0.71
     pim
    -0.70
     churn
    -0.69
     vag
    -0.69
     langu
    -0.68
     tongues
    -0.68
     offending
    -0.66
     limestone
    -0.66
     temptation
    -0.65
    POSITIVE LOGITS
     Accountability
    0.97
     Equality
    0.95
     Progress
    0.94
     Respons
    0.89
     Peace
    0.88
     Inquiry
    0.82
     Action
    0.81
     Democracy
    0.81
     Investigative
    0.81
     Sustainable
    0.80
    Activations Density 0.018%

    No Known Activations