© 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 24576
    9. 8-RES_FS24576-JB
    11. 1701
    Prev
    Next
    INDEX
    Explanations

    phrases related to detailed explanations and information

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs24576-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    24,576
    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
    195
    +0.11
    0.5%
    302
    +0.10
    0.4%
    497
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    195
    +0.11
    0.01
    302
    +0.10
    0.02
    117
    +0.09
    0.02
    Negative Logits
    leader
    -0.76
    CENT
    -0.76
    earth
    -0.75
    hop
    -0.74
    acus
    -0.73
    Earth
    -0.73
    instead
    -0.72
    hops
    -0.71
    leaders
    -0.69
    fortunately
    -0.69
    POSITIVE LOGITS
     amounts
    0.97
    mble
    0.88
     renovations
    0.81
     documentation
    0.78
     schooling
    0.78
     tracts
    0.77
     customization
    0.77
     quantities
    0.77
     coverage
    0.75
     detail
    0.74
    Activations Density 0.045%

    No Known Activations