© 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-2-27B
    3. 10-GEMMASCOPE-RES-131K
    4. 19336
    Prev
    Next
    INDEX
    Explanations

    trityl oxide

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

    abbreviations or acronyms that consist of 2-3 capital letters, often followed by numeric or symbolic characters.

    oai_token-act-pair · claude-3-7-sonnet-20250219Triggered by @neilrathi

    The neuron is primarily activated by all-caps letter sequences (acronyms or initialisms).

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_10/width_131k
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    is
    -1.79
    ing
    -1.64
    atguigu
    -1.57
    𖥧
    -1.55
     NavController
    -1.48
     cercana
    -1.45
    都没有
    -1.42
     attaining
    -1.41
    겁
    -1.40
     típica
    -1.38
    POSITIVE LOGITS
     sowohl
    1.70
     zarówno
    1.54
     have
    1.54
     далее
    1.50
    蠵
    1.47
     появились
    1.45
    ANCE
    1.43
    ZING
    1.42
    ЖИ
    1.40
     vocale
    1.40
    Activations Density 0.110%

    No Known Activations