© 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. 14-GEMMASCOPE-2-TRANSCODER-262K
    4. 98580
    Prev
    Next
    INDEX
    Explanations

    **Analyzing Neuron Behavior**I am currently processing the provided lists to discern patterns that might represent the activation of a neuron. My goal is to identify what the neuron signals by finding commonalities among the inputs.**MAX_ACTIVATING_TOKENS**:'eruption', 'eruptions', 'canic', 'Mount', 'eruptions', 'eruptions', 'volcanic', 'eru'**TOKENS_AFTER_MAX_ACTIVATING_TOKEN**:'s', '1', 'released', ':**', 'winter', 'St', 'contribute', 'eruptions', 'ptions'**TOP_POSITIVE_LOGITS**:'Lemb', 'స్‌', 'cles', 'institut', 'аз', 'ri', 'lié', 'oes', 'ésia', 'au'**TOP_ACTIVATING_TEXTS**:"most significant volcanic eruptions, and precise by which metric you've assessed their significance""top 5 most significant volcanic eruptions, ranked by **Volcanic Explosivity Index (VEI""the 1980 Mount St. Helens eruption released about 1 cubic kilometer of material.""**Volcanic eruptions:** Release ash and gases that form PM.""The supervolcano Toba (in present-day Indonesia) erupted with massive force, causing a "volcanic winter" that lasted for years...""The eruption isn’t a classic, explosive event like Mount St. Helens. Instead, it’s characterized by a series of phreatic eruptions – steam-driven explosions caused""Activity:** Volcanoes release greenhouse gases (like CO2) and aerosols (tiny particles). *Argument:* Volcanic eruptions contribute to climate change.""b) Burning of fossil fuels c) Volcanic eruptions d) Agricultural practices""Massive earthquakes, volcanic eruptions, a meteor strike, a sudden climate shift.""Create land, impact climate, and can be dangerous (eruptions, mudflows, ashfall)."**Pattern Identification**:The `MAX_ACTIVATING_TOKENS` heavily feature "eruption", "eruptions", and "volcanic".The `TOP_ACTIVATING_TEXTS` consistently discuss "volcanic eruptions", "eruptions", and specific examples like "Mount St. Helens eruption".The `TOKENS_AFTER_MAX_ACTIVATING_TOKEN` include follow-ups like "s" (for plural "eruptions"), "released" (describing an eruption), "winter" (following "volcanic" for "volcanic winter"), and "contribute" (describing effects of eruptions).The common theme is the concept of volcanic eruptions and their effects or descriptions. The `TOP_POSITIVE_LOGITS` list seems unrelated and possibly noise or from a different context, so it's being discounted for finding the primary pattern.**Explanation**: The pattern revolves around the word "eruption" and its association with volcanoes, their effects, and specific instances.**Final Answer Derivation**:* MAX_ACTIVATING_TOKENS: 'eruption', 'eruptions', 'volcanic'* TOP_ACTIVATING_TEXTS: "volcanic eruptions", "eruptions", "Mount St. Helens eruption", "volcanic winter"The neuron is clearly activated by discussions of volcanic eruptions.**Concise Explanation**: volcanic eruptions

    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_14_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
    用戶
    0.52
    有很多
    0.52
    用户信息
    0.47
    波
    0.46
     terrorism
    0.46
    blade
    0.46
    worst
    0.46
     favored
    0.46
    文档
    0.44
    γη
    0.44
    POSITIVE LOGITS
     Lemb
    0.52
    స్‌
    0.50
    cles
    0.49
    institut
    0.49
    аз
    0.49
    ri
    0.48
     lié
    0.48
    oes
    0.47
    ésia
    0.47
     au
    0.46
    Activations Density 0.001%

    No Known Activations