© 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
    3. Gemma-3-27B-IT
    5. 37-GEMMASCOPE-2-TRANSCODER-262K
    7. 138795
    Prev
    Next
    INDEX
    Explanations

    - **TOP_ACTIVATING_TEXTS**: - Text 1: Mentions Cyrillic character mappings like 'ћ': 'ć', 'ь': '', 'ъ': ''. Describes palatalized consonants 'ć/ś/ź/ń' as "softer, often with a 'y' sound following." - Text 2: A user asking for Polish words consisting of specific Polish letters (ą, ć, ę, ł, ń, ś, ó, ż, ź). The model response starts with "Okay," and then lists Cyrillic mappings again, including 'ь': ''. - Text 3: A user asking about the encoding of "🇨🇦". The model response includes Cyrillic mappings like 'А' and '

    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_37_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
     pegs
    0.40
     Lynd
    0.39
     What
    0.38
     berry
    0.38
     pegged
    0.38
     Mandy
    0.38
    ້ນ
    0.37
     Miro
    0.36
    奻
    0.36
    養
    0.36
    POSITIVE LOGITS
    alaman
    0.38
    менте
    0.38
    وله
    0.37
    aze
    0.36
    FormComponent
    0.35
    idE
    0.35
     Coca
    0.35
    
    0.35
    agram
    0.34
     mutil
    0.34
    Activations Density 0.001%

    No Known Activations