© 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-3-12B
    5. 31-GEMMASCOPE-2-RES-16K
    7. 1845
    Prev
    Next
    INDEX
    Explanations

    "hello" string literal

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

    The neuron is essentially picking out extremely common short tokens—mostly high-frequency function words (like “by,” “and,” “in,” small two‐letter pieces, or punctuation/whitespace).

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-12b-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
     Spur
    0.62
     Pledge
    0.61
     Platz
    0.60
     Vanden
    0.59
     pot
    0.59
    Van
    0.56
    ર
    0.56
     Venue
    0.56
    Sp
    0.55
     Van
    0.55
    POSITIVE LOGITS
    ṡ
    0.74
     méxico
    0.70
    msgs
    0.67
    gpu
    0.66
    owego
    0.66
    ູ່
    0.65
    phdr
    0.65
     приро
    0.65
    umd
    0.64
    ʌ
    0.64
    Activations Density 0.633%

    No Known Activations