© 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
    2. Gemma-2-27B
    3. 22-GEMMASCOPE-RES-131K
    4. 95337
    Prev
    Next
    INDEX
    Explanations

    special attention

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

    The neuron fires strongly on words from formal documentation or guideline excerpts—especially content‐heavy technical or policy phrases (e.g. “enterprise,” “class,” “database,” “point,” “time,” “recovery,” “avoid,” “asking,” “subjective,” etc.). It’s essentially spotting jargon‐laden, documentation-style noun phrases.

    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
    èlement
    -1.00
     solch
    -0.99
     vícti
    -0.98
    Seni
    -0.92
     perdagangan
    -0.88
    ۟
    -0.88
     گا
    -0.88
    斯的
    -0.87
     fometimes
    -0.87
     haupts
    -0.86
    POSITIVE LOGITS
     colgantes
    0.99
     universitaire
    0.96
    Although
    0.93
     ...
    0.92
     religieux
    0.92
     …
    0.91
     Procu
    0.88
    During
    0.88
    Others
    0.88
    ө
    0.87
    Activations Density 0.054%

    No Known Activations