© 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 EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-27B-IT
    3. 11-GEMMASCOPE-2-TRANSCODER-262K
    4. 40314
    Prev
    Next
    INDEX
    Explanations

    neuron explains biological feeding and life cycle behaviors, focusing on insects and plants.The `MAX_ACTIVATING_TOKENS` list contains nouns and verbs related to how organisms interact with their environment and each other, especially in a biological context:- `sap`: implies feeding behavior, often from plants.- `sition`: part of `oviposition` (egg-laying).- `plant`: a common environment or food source.- `parts`: could refer to body parts involved in feeding or reproduction.- `host`: implies parasitism or symbiotic relationships.- `Feeding`: a core biological process.- `leaf`: part of a plant, habitat for insects, or part of their diet.- `Egg`: a life stage.- `probe`: a method of feeding or sensing.- `osc`: could be part of `nectar` or `proboscis`.The `TOKENS_AFTER_MAX_ACTIVATING_TOKEN` list provides context following these terms:- `from`- `juices` (e.g., plant juices)- `look`- `is`Combining these, the neuron seems to be activated by discussions of **biological feeding habits and life stages, particularly concerning plants and insects**.Let's refine this:- `sap`, `plant`, `juices` -> feeding on plants.- `Egg`, `oviposition` (from `sition`) -> life cycle.- `Feeding`, `probe`, `host` -> feeding mechanisms and relationships.The `TOP_POSITIVE_LOGITS` list includes things like `il`, `Saint`, `acking`, `eng`, `riage`, `Eight` which seem less directly related but might indicate specific contexts or unusual words that activate it positively when combined with the core pattern.insect feeding, egg laying, plant juices

    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_11_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
     मिली
    0.56
     butadiene
    0.55
    LOR
    0.50
    ్ఞ
    0.50
    Якщо
    0.50
     anem
    0.50
     ajj
    0.49
     hydraz
    0.48
     Dons
    0.48
    DOT
    0.47
    POSITIVE LOGITS
    il
    0.53
    ፅ
    0.51
     Saint
    0.48
    acking
    0.46
    eng
    0.45
    riage
    0.45
    ográfica
    0.45
    ጽ
    0.45
     Eight
    0.45
    individual
    0.44
    Activations Density 0.022%

    No Known Activations