© 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-3-27B-IT
    3. 46-GEMMASCOPE-2-TRANSCODER-262K
    4. 204528
    Prev
    Next
    INDEX
    Explanations

    Finding patterns in lists.MAX_ACTIVATING_TOKENS: structural elements.TOKENS_AFTER_MAX_ACTIVATING_TOKEN: specific nouns.TOP_POSITIVE_LOGITS: specific words, names, concepts.TOP_ACTIVATING_TEXTS: descriptive of specific domains or entities.The neuron seems to be triggered by specific technical terms, names, or defined concepts presented in the text. If the neuron is looking for *specific* items, what's a concise way to say it?Let's try to relate it to the examples:'generate', 'model', 'group' (Method 2) are specific nouns.'Gideon', 'War', 'Ge

    np_max-act-logits · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-27b-it/transcoder_all/layer_46_width_262k_l0_small_affine
    Prompts (Dashboard)
    238,145 prompts, 512 tokens each
    Dataset (Dashboard)
    lmsys + oasst1
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    цтва
    0.44
     Pari
    0.39
    dex
    0.37
     kr
    0.36
    এক্স
    0.36
    Imp
    0.36
    Bub
    0.36
    kr
    0.35
    CharAt
    0.35
    MP
    0.35
    POSITIVE LOGITS
     Gideon
    0.47
     Greedy
    0.40
    丟
    0.40
     আক্রমন
    0.39
     War
    0.39
    ichts
    0.39
     Geek
    0.39
    ൮
    0.39
     বালক
    0.38
     Way
    0.38
    Activations Density 0.000%

    No Known Activations