© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it
    Neuronpedia
    Jacobian LensNEW
    Natural Language
    Autoencoders
    MAY 26
    Assistant AxisJAN 26Circuit TracerAPR 26SteerSAE EvalsExportsinterp-engineNEW Community BlogPrivacy & TermsContact
    Side-by-side compare needs a wider screen.
    Open other featureClose compare
    PrevNext
    Close
    Circuit Sparsity
    No connected neurons data available
    Explanations

    try block start (negative activations)

    python-code · gemini-2.5-flash-lite
    Auto-Interp Tools
    Top Features by Cosine Similarity
    Configuration
    Prompts (Dashboard)
    16,384 prompts, 256 tokens each
    Dataset (Dashboard)
    neuronpedia/python-code-simplified
    No Configuration Found
    Embeds
    IFrame
    Link
    Negative Logits
     continue
    -0.08
    ?
    -0.07
    _start
    -0.06
    ype
    -0.05
    _end
    -0.05
    (start
    -0.05
     ny
    -0.04
    a
    -0.04
    b
    -0.04
    start
    -0.04
    Positive Logits
    ative
    0.07
    .heapp
    0.07
    ion
    0.06
    ance
    0.06
    ied
    0.06
    ren
    0.06
    utes
    0.05
    istic
    0.05
    alue
    0.05
    ations
    0.05
    Activations Density 79.000%

    No Known Activations

    PrevNext
    Close
    Circuit Sparsity
    No connected neurons data available
    Explanations

    comple★x★return comple★x★(f, g)f = b.real + c.realg = b.imag + c.imagreturn comple★x★(f, g)The neuron seems to be detecting the `complex` type/keyword in Python, especially when it's part of a function call like `complex(real_part, imag_part)`.complex number constructor (negative activations)

    python-code · gemini-2.5-flash-lite
    Auto-Interp Tools
    Top Features by Cosine Similarity
    Configuration
    Prompts (Dashboard)
    16,384 prompts, 256 tokens each
    Dataset (Dashboard)
    neuronpedia/python-code-simplified
    No Configuration Found
    Embeds
    IFrame
    Link
    Negative Logits
    .nodes
    -0.10
    .remove
    -0.09
    .node
    -0.09
    .position
    -0.09
    .heapp
    -0.09
    [node
    -0.08
    §
    -0.07
    .items
    -0.07
    ep
    -0.07
    cell
    -0.07
    Positive Logits
     instruction
    0.10
     instructions
    0.09
    p
    0.09
    struction
    0.08
     command
    0.08
    structions
    0.07
     per
    0.07
    gram
    0.07
    per
    0.06
    mory
    0.06
    Activations Density 99.248%

    No Known Activations