© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-2-27B
    5. 22-GEMMASCOPE-RES-131K
    7. 128088
    Prev
    Next
    INDEX
    Explanations

    This neuron responds to programming‐style tokens—e.g. code commands (like DO, &&, rd), punctuation or path separators (/, “?”), macro/keyword identifiers, and numeric literals—rather than ordinary words.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408

    technical parameters and programming constructs

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_22/width_131k
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    撫
    -0.96
     fortsatt
    -0.93
    iree
    -0.90
    setDo
    -0.89
    earch
    -0.86
     detectives
    -0.85
    DoubleClick
    -0.85
     brukes
    -0.85
    }(
    -0.84
     legado
    -0.84
    POSITIVE LOGITS
    こちらも
    1.14
     asimismo
    1.05
    これも
    1.02
     illumin
    0.96
    Deserialize
    0.96
    tableWidget
    0.94
     dinam
    0.94
     inscri
    0.92
    さまざま
    0.92
     sogar
    0.90
    Activations Density 0.008%

    No Known Activations