© 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. 16-GEMMASCOPE-2-TRANSCODER-262K
    4. 217249
    Prev
    Next
    INDEX
    Explanations

    The pattern seems to revolve around concepts of **"functions"**, **"dimensions"**, and **"reversing/opposite order"** (like in the word reversal example). The Korean word '감' (gam) appears in the activating tokens, often related to Korean expressions of feeling or thanks ('감사합니다', '감사해요').Let's look for overlaps and themes.* **Functions**, **Functional**, **Useful**: Suggests operations, purpose, utility.* **Dimensions**: Relates to abstract properties, aspects, or ways of measuring.* **Reversing a word**: Implies an operation on sequences, symmetry, or inverse operations.* **감 (gam)**: Can relate to feelings or thanks. In the context of 'warmer than 감사합니다', it points to nuanced emotional expression.The `TOP_POSITIVE_LOGITS` lean heavily towards programming/computational terms like "Functions", "Anaconda", "Functional", "condensation".The `MAX_ACTIVATING_TOKENS` are a mix: 'or', '감', 'represents', 'H', 'NESS', '감', '😂', 'soup', 'come'.The `TOKENS_AFTER_MAX_ACTIVATING_TOKEN` are: 'book', '사', 'a', ';', 'A', '사', 'of', 'from', 'Here'.**functions and abstract concepts**

    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
     Newell
    0.51
     medicina
    0.49
     nostra
    0.48
    nesto
    0.47
     наш
    0.46
     cardNumber
    0.46
     Reggie
    0.46
     Reg
    0.46
    nih
    0.46
    masına
    0.45
    POSITIVE LOGITS
    Functions
    0.54
    Anaconda
    0.54
    condensation
    0.53
    Function
    0.53
    Useful
    0.53
    M
    0.51
    ("
    0.51
     Functions
    0.50
    Relative
    0.50
    Functional
    0.50
    Activations Density 0.000%

    No Known Activations