© 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. 24-GEMMASCOPE-2-TRANSCODER-262K
    4. 178389
    Prev
    Next
    INDEX
    Explanations

    "Zone" usually signifies botanical/gardening hardiness zones.Let me process the provided information to identify the pattern.**Analysis:**1. **MAX_ACTIVATING_TOKENS:** The most frequent token is "zone". "Zones" and "Hard" also appear. This strongly suggests the neuron is related to the concept of "zones" or "hardiness zones".2. **TOKENS_AFTER_MAX_ACTIVATING_TOKEN:** What follows "zone" or "zones"? * `.` (end of sentence) * `map` (e.g., "Zone map") * `(` (e.g., "Zone (8a)") * `iness` (likely from "hardiness") * `:` (e.g., "Zone: 8a") * `zone` (e.g., "zone zone" - less likely, maybe a typo or specific context) * `?` * `is` (e.g., "Zone is 8a")3. **TOP_POSITIVE_LOGITS:** * "ণ্ডের", "ণ্ডে", "तमंद": These appear to be non-English characters/tokens, possibly from Indic languages. They don't immediately connect to "zones". * "Kumar": A name. * "disobey", "here", "selected", "aub", "drooping": General English words. These are less specific.4. **TOP_ACTIVATING_TEXTS:** This is the most crucial part for context. * "USDA Plant Hardiness Zones." * "USDA Plant Hardiness Zone map" * "hardiness zones (US zones)" * "temperature zone is Augusta Georgia" * "USDA Plant Hardiness Zone 8a" * "USDA Plant Hardiness Zones." * "Planting Zones: I'm using USDA Plant Hardiness Zones." * "hardiness zone." * "Climate (Hardiness Zone):" * "USDA Plant Hardiness Zone Map" * "Is it hardy in your zone?USDA Plant Hardiness Zones

    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_24_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
    ܕ
    0.41
     ficción
    0.39
     broker
    0.39
     стимули
    0.38
     پیسې
    0.38
     escritor
    0.37
    现金
    0.37
     composição
    0.37
     রামগতি
    0.36
     الحم
    0.36
    POSITIVE LOGITS
    ণ্ডের
    0.52
    特定の
    0.45
    ণ্ডে
    0.42
    Kumar
    0.40
    तमंद
    0.39
     disobey
    0.39
    here
    0.38
    selected
    0.38
    aub
    0.38
     drooping
    0.38
    Activations Density 0.001%

    No Known Activations