© 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. 10312
    Prev
    Next
    INDEX
    Explanations

    mentions of specific actions or activities in combination

    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
    476
    +0.12
    0.5%
    738
    +0.11
    0.5%
    471
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    172
    +0.12
    0.04
    476
    +0.11
    0.04
    738
    +0.10
    0.04
    Negative Logits
    elled
    -0.82
    oux
    -0.81
    ettle
    -0.80
    ology
    -0.77
    hack
    -0.77
    iotic
    -0.75
    aer
    -0.75
    ological
    -0.74
    omorphic
    -0.74
    ulous
    -0.74
    POSITIVE LOGITS
     maintaining
    1.13
     researching
    1.09
     awaiting
    1.08
     retaining
    1.08
     preserving
    1.05
     simultaneously
    1.01
     touring
    1.00
     avoiding
    0.99
     traveling
    0.96
     minimizing
    0.96
    Activations Density 0.081%

    No Known Activations