© 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. 31-GEMMASCOPE-2-TRANSCODER-262K
    4. 76571
    Prev
    Next
    INDEX
    Explanations

    **TOP_ACTIVATING_TEXTS**:- "If the foreign corporate entity owns a commercial property..."- "This means *if you owe someone money, you can legally offer them cash..."*- "*If the person *uses* the fake will..."*- "*If the buyer backs out of a house sale..."*- "*If a suspension is due to a clear violation of the ToS..."*- "*If your license has been inactive for 5 years or less..."*- "*If your salary meets this, you are eligible."*- "*If the *aggregate* value of *all* your foreign financial accounts..."*- "*If the parents are both, for example, US citizens, the child would likely be considered a US citizen..."*- "*If a site advertises an onion service, Tor Browser will *automatically* redirect you..."*The dominant pattern is the word "If" followed by a condition or a clause that describes a situation, often leading to a consequence or a statement of fact. The `MAX_ACTIVATING_TOKENS` being "If" supports this.if condition leads to statement

    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_31_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
     if
    0.50
     jei
    0.40
     إذا
    0.38
     jika
    0.38
    if
    0.37
     اذا
    0.37
     если
    0.37
     nếu
    0.37
     якщо
    0.36
     wenn
    0.35
    POSITIVE LOGITS
     ένας
    0.40
     واحدة
    0.36
    ាយ
    0.33
     непотпу
    0.33
    移植
    0.33
     하나의
    0.33
     पहाड़
    0.33
     чолові
    0.32
     వారి
    0.32
    넛
    0.32
    Activations Density 0.008%

    No Known Activations