© 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. 62151
    Prev
    Next
    INDEX
    Explanations

    useruserUseruserUseruserпользовательпользовательuserUseruser</MAX_ACTIVATING_TOKENS><TOKENS_AFTER_MAX_ACTIVATING_TOKEN>виские<end_of_turn>saysquestion,ватисяpostscan_</TOKENS_AFTER_MAX_ACTIVATING_TOKEN><TOP_POSITIVE_LOGITS>through通过THROUGH为zoomprzezผ่านz思っśnie</TOP_POSITIVE_LOGITS><TOP_ACTIVATING_TEXTS>isActive]="true"> <span title>Участник</span> <span desc> Данный пользователь видит все проекты платформы и всех участников, может написать кому угодно и предложить сотрудничество, владеет настройки внешнего вида и функциональности. Можно менять лаунчеры, виджеты, темы и даже устанавливать пользовательские прошивки. * **Открытый исходный код:** Это позволяет разработчикам создавать более Answer with an understanding of the meaning and using a modern style of communication, and "briefly" as a normal user<end_of_turn> <start_of_turn>model Whoa, that's *intense*! Seriously, what you're describing –00% of the following statements: 1. 1+1=2 2. In the last message User says "connect with human" 3. In the last message User asks about order information" 4. In theozhnaya Street, d7". The company sells servers and works from 10 to 17. User question: "Hello. what is the work schedule? And where are you located?" Answer:<end_of_turn> <start_of_turn>model * "Меня завораживает процесс от идеи до реализации. Начиная с исследования потребностей пользователей, продумывания сценавання спокійної музики. * **Обмежте екранний час:** Припиніть користуватися телефонами, планшетами, комп'ютерами та телевізором принаймні за годи crucial for improving the accuracy and relevance of subsequent analysis and predictive modeling. Raw text data from VKontakte, consisting of user posts, comments, and messages, is inherently noisy and contains significant morphological variation. Applying lemmatization reduces words to theirter. User can create observations spots of animals with geolocation and pictures. Types of animals in second table. User can search by animals, countries and spots.<end_of_turn> <start_of_turn>model Okay, here's a comprehensive PostgreSQL schema andcursor.execute(""" CREATE TABLE IF NOT EXISTS reminders ( id INTEGER PRIMARY KEY AUTOINCREMENT, user_id INTEGER NOT NULL, reminder_text TEXT NOT NULL, reminder_time TEXT NOT NULL </TOP_ACTIVATING_TEXTS>Explain the neuron above with a word or phrase, not a complete sentence

    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
     التيار
    0.59
    ாதீர்கள்
    0.54
     Futuristic
    0.54
    нары
    0.52
    න්ධ
    0.52
    椹
    0.52
     Dynamo
    0.52
    ната
    0.51
    ЕТ
    0.51
     Ministério
    0.50
    POSITIVE LOGITS
     through
    0.63
     通过
    0.60
     THROUGH
    0.59
     为
    0.58
     zoom
    0.57
     przez
    0.57
     ผ่าน
    0.57
     z
    0.57
     思っ
    0.56
    śnie
    0.55
    Activations Density 0.000%

    No Known Activations