© 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. 34-GEMMASCOPE-RES-131K
    4. 102763
    Prev
    Next
    INDEX
    Explanations

    rendered safe or innocuous

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

    The neuron strongly activates on very short, high‐frequency tokens (especially two‐ or three‐letter function words like “to,” “in,” “de,” “be,” “a,” etc.).

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_34/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
    NIP
    -0.79
    แล
    -0.71
    生活的
    -0.71
    nal
    -0.68
    atience
    -0.68
     chắn
    -0.67
    سور
    -0.67
    čenje
    -0.67
     esmeralda
    -0.66
    阴阳
    -0.66
    POSITIVE LOGITS
     harmless
    6.19
     innoc
    3.72
     benign
    3.48
    innoc
    2.28
     safe
    2.25
     innocent
    2.22
     безопас
    2.06
     безо
    1.98
     Innoc
    1.97
    Innoc
    1.91
    Activations Density 0.124%

    No Known Activations