© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW
    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-27B
    3. 31-GEMMASCOPE-2-RES-16K
    4. 16167
    Prev
    Next
    INDEX
    Explanations

    cardinal numbers**Rationale:**The `TOP_POSITIVE_LOGITS` list is dominated by cardinal numbers (thirteen, seventeen, eighteen, nineteen, eighty, fourteen, twelve). While other tokens in `MAX_ACTIVATING_TOKENS` and `TOP_ACTIVATING_TEXTS` relate to travel, books, and specific names, the overwhelming signal in the logits points towards numerical concepts. The phrase "cardinal numbers" is concise and accurately reflects this primary signal

    np_acts-logits-general · gemini-2.5-flash-lite

    pronoun + verb

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-27b-pt/resid_post/layer_31_width_16k_l0_medium
    Prompts (Dashboard)
    392,802 prompts, 256 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    倘
    0.66
    當中
    0.63
    Asimismo
    0.59
     notamment
    0.57
    Comme
    0.57
     лишь
    0.57
    няка
    0.56
    👌
    0.56
    แม้
    0.55
     désormais
    0.54
    POSITIVE LOGITS
     thirteen
    0.63
     seventeen
    0.61
     eighteen
    0.60
     nineteen
    0.60
     zuerst
    0.58
     dört
    0.58
     eighty
    0.58
     fourteen
    0.57
     இரண்டு
    0.55
     twelve
    0.52
    Activations Density 0.041%

    No Known Activations