© 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 EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-27B-IT
    3. 16-GEMMASCOPE-2-TRANSCODER-262K
    4. 214923
    Prev
    Next
    INDEX
    Explanations

    The neuron seems to be associated with explanations that involve the 'causes' and 'effects' of events or concepts, particularly historical or strategic ones. The presence of "revolution" and "safety" suggests it might be identifying explanations of processes that lead to significant outcomes or ensure well-being. The example texts show it activating for explanations of historical events ("american revolution"), consequences of actions ("assohole"), and operational safety ("safety").Let's look at the tokens again.`MAX_ACTIVATING_TOKENS`: `and`, `asso`, `suggest`, `is`, `revolution`, `safety`, `might`, `enna`, `anding`, `dir``TOKENS_AFTER_MAX_ACTIVATING_TOKEN`: `interests`, `hole`, `a`, `currently`, `<end_of_turn>`, `.`, `be`, `and`, `:**`, `,``TOP_POSITIVE_LOGITS`: `Healthy`, `Think`, `ف`, `Online`, `Oracle`, `Wish`, `Live`, `Safe`, `ى`, `Great``TOP_ACTIVATING_TEXTS`:- "All property and interests in property that hereafter come within the United States of the following persons are blocked" - `interests` after `and`- "you're an assohole and an incompetent. you're fired . Never want to see you again" - `asso` is in `MAX_ACTIVATING_TOKENS`, `hole` is after `asso` in `assohole`.- "american revolution" - `revolution` is in `MAX_ACTIVATING_TOKENS`.- "I'll also cover safety." - `safety` is in `MAX_ACTIVATING_TOKENS`.causes and effects of revolution safety

    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_16_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
     seeker
    0.53
    atement
    0.49
     wink
    0.49
     leaching
    0.48
     paraffin
    0.47
     quirk
    0.47
    ides
    0.46
    atically
    0.46
     dimensional
    0.46
     coffin
    0.46
    POSITIVE LOGITS
    Healthy
    0.53
    Think
    0.48
    ف
    0.48
    Online
    0.46
    Oracle
    0.46
    Wish
    0.46
    Live
    0.46
    Safe
    0.46
    ى
    0.46
    Great
    0.44
    Activations Density 0.000%

    No Known Activations