Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    © Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Feature Splitting for GPT2-Small
    3. GPT2-Small
    4. Residual Feature Splitting 12288
    5. 8-RES_FS12288-JB
    6. 277
    Prev
    Next
    INDEX
    Explanations

    mentions of large numerical quantities like 'hundreds' or 'thousands'

    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
    497
    +0.16
    0.8%
    71
    +0.11
    0.5%
    718
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    497
    +0.16
    0.04
    71
    +0.11
    0.03
    129
    +0.10
    0.03
    Negative Logits
    agate
    -0.64
     hegemony
    -0.61
     Redemption
    -0.60
    vice
    -0.58
     advice
    -0.58
     goodness
    -0.57
    inventoryQuantity
    -0.57
     Situation
    -0.57
     pessim
    -0.56
     Ital
    -0.55
    POSITIVE LOGITS
    ourcing
    1.09
    paces
    1.04
     thousand
    0.98
    ourced
    0.95
    entimes
    0.84
    chool
    0.82
    ousands
    0.81
     thousands
    0.80
    eenth
    0.80
     of
    0.79
    Activations Density 0.076%

    No Known Activations