© 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. 15-GEMMASCOPE-2-TRANSCODER-262K
    4. 16376
    Prev
    Next
    INDEX
    Explanations

    This neuron is strongly associated with the word "deck" and its variations. Looking at the `TOKENS_AFTER_MAX_ACTIVATING_TOKEN`, we see common continuations like "and", "off", and words related to cleaning ("cleaner") or structural elements ("jo"). The `TOP_ACTIVATING_TEXTS` reinforce this by providing contexts like "build a 10x12 foot deck", "Deck & Patio Cleaning", "preferred method for deck joists", "used for decks or patios", "Wood Decking", "beautiful decks", "Use a deck cleaner", "under decks and porches". All these suggest the neuron is activated by discussions related to outdoor wooden structures, particularly decks.The `TOP_POSITIVE_LOGITS` seem to be an artifact of a multilingual model or a specific training setup, potentially unrelated to the core semantic meaning associated with "deck" in English texts. The task is to find a pattern *within* the English activating texts and tokens.The dominant pattern is the word "deck" and phrases or contexts where it appears, often associated with construction, maintenance, or placement.Therefore, a concise explanation would be:"deck"

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-27b-it/transcoder_all/layer_15_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
     Childhood
    0.50
     एंड्रॉयड
    0.49
    ЕТ
    0.49
    settes
    0.49
     والدہ
    0.49
    migration
    0.48
     Centrale
    0.47
    scp
    0.47
     Learning
    0.47
     নয়
    0.47
    POSITIVE LOGITS
    をお
    0.46
     truyền
    0.45
    እ
    0.44
    ఉ
    0.43
     mehreren
    0.43
     pemb
    0.43
     xm
    0.43
    
    0.43
     lia
    0.42
     fasci
    0.42
    Activations Density 0.001%

    No Known Activations