© 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. 31-GEMMASCOPE-2-TRANSCODER-262K
    4. 163112
    Prev
    Next
    INDEX
    Explanations

    analysis of the lists:- **MAX_ACTIVATING_TOKENS**: `, ' ' , in Inc by . ,` - these are mostly punctuation or common short words.- **TOKENS_AFTER_MAX_ACTIVATING_TOKEN**: `only s They s vic the as killing the` - I see "only", "vic", "killing".- **TOP_POSITIVE_LOGITS**: `夺 回 सल Tread friday inform enang 奪 comté शुक्रवार` - this list has a mix of non-English words (Hindi for Friday, Chinese characters) and English words ("friday", "inform", "comté").- **TOP_ACTIVATING_TEXTS**: The texts heavily mention "Inca royalty", "vicuña wool", "fiber", historical context, "conservation status", "significance". One text discusses "sacrificed by ritual". Another mentions "killing the animal" for harvesting.Looking for common themes and patterns:1.Inca royalty, vicuña, fiber

    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
     Peruvian
    0.46
     Cassini
    0.44
     Peru
    0.42
     Bolivia
    0.41
     Requirement
    0.41
     Radioactive
    0.40
     Zamb
    0.38
     cassava
    0.38
     Amazon
    0.37
     postulates
    0.37
    POSITIVE LOGITS
    夺
    0.47
    回
    0.44
    सल
    0.41
     Tread
    0.39
     friday
    0.39
    inform
    0.38
    enang
    0.38
    奪
    0.38
     comté
    0.37
     शुक्रवार
    0.37
    Activations Density 0.006%

    No Known Activations