© 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 49152
    9. 8-RES_FS49152-JB
    11. 37348
    Prev
    Next
    INDEX
    Explanations

    questions starting with "how" or "why"

    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
    218
    +0.13
    0.6%
    679
    +0.13
    0.6%
    456
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    456
    +0.13
    0.02
    449
    +0.13
    0.02
    287
    +0.11
    0.02
    Negative Logits
    thur
    -0.73
    asant
    -0.71
    cedented
    -0.71
    wcsstore
    -0.69
    agonists
    -0.68
    heter
    -0.67
    dayName
    -0.67
     advoc
    -0.66
    esville
    -0.66
    itud
    -0.66
    POSITIVE LOGITS
    actic
    0.77
    ?]
    0.75
     happen
    0.67
     someone
    0.65
     astronomers
    0.64
    ={
    0.64
     emerge
    0.64
     somebody
    0.63
     evolve
    0.62
     mathematic
    0.62
    Activations Density 0.019%

    No Known Activations