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

    Neuronpedia

    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-2-27B
    5. 22-GEMMASCOPE-RES-131K
    7. 95557
    Prev
    Next
    INDEX
    Explanations

    Computed tomography

    np_acts-logits-general · gemini-2.5-flash-lite

    error handling code blocks in programming languages.

    oai_token-act-pair · claude-3-7-sonnet-20250219Triggered by @neilrathi

    The neuron activates on exception‐handling constructs (e.g. “catch,” “error,” logging calls), i.e. it detects error‐handling code.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    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
     Légion
    -1.13
     compositeur
    -1.08
     bruise
    -1.08
    蔦
    -1.06
     düğün
    -1.05
     Sinf
    -1.02
    鯵
    -0.99
    鰯
    -0.99
     电子
    -0.98
     загряз
    -0.98
    POSITIVE LOGITS
     never
    1.08
     came
    1.07
     or
    1.00
     described
    0.94
     just
    0.90
     all
    0.88
     would
    0.88
     simply
    0.87
     here
    0.86
     took
    0.84
    Activations Density 0.002%

    No Known Activations