© 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. 1637
    Prev
    Next
    INDEX
    Explanations

    the phrase "X and a half"

    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
    45
    +0.10
    0.5%
    610
    +0.09
    0.4%
    363
    +0.08
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    610
    +0.10
    0.03
    716
    +0.09
    0.02
    45
    +0.08
    0.02
    Negative Logits
    NSA
    -0.72
    ACTION
    -0.62
    NAS
    -0.62
    REP
    -0.60
    TEXTURE
    -0.58
    BALL
    -0.58
     Aber
    -0.57
    UTF
    -0.57
    Reloaded
    -0.56
    Enhanced
    -0.56
    POSITIVE LOGITS
     half
    0.78
    half
    0.74
     thirds
    0.71
     couple
    0.69
    oba
    0.68
     dime
    0.64
     few
    0.64
     handful
    0.64
    ixty
    0.63
    oes
    0.63
    Activations Density 0.044%

    No Known Activations