© 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. 9243
    Prev
    Next
    INDEX
    Explanations

    the word "our."

    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
    385
    +0.12
    0.5%
    301
    +0.11
    0.5%
    152
    +0.11
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    232
    +0.12
    0.01
    304
    +0.11
    0.01
    428
    +0.11
    0.01
    Negative Logits
     steroids
    -0.68
    arlane
    -0.67
     Rasmussen
    -0.64
     relapse
    -0.63
    ORN
    -0.61
    aeda
    -0.61
    ellectual
    -0.61
     multic
    -0.61
     duplication
    -0.59
     buckle
    -0.59
    POSITIVE LOGITS
    neau
    1.30
    dain
    1.19
    selves
    1.19
    neys
    1.08
    cery
    1.07
    ishment
    1.00
    ¯¯¯¯¯¯¯¯
    0.94
    ishing
    0.93
    tesy
    0.92
    din
    0.92
    Activations Density 0.020%

    No Known Activations