© 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-3-12B
    3. 31-GEMMASCOPE-2-RES-16K
    4. 8549
    Prev
    Next
    INDEX
    Explanations

    photorealistic, autonomy

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

    The neuron consistently fires on the initial subword fragments (three‐ or four‐letter prefixes) at the beginnings of longer, multi‐syllable words.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408

    partial word fragments that are part of compound words or technical terms, particularly focusing on prefixes and word boundaries.

    oai_token-act-pair · claude-4-5-sonnetTriggered by @jyhe0408

    prefix-like word stems at the start of longer, often technical or proper-noun terms.

    oai_token-act-pair · gpt-5Triggered 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
     aveva
    0.75
     veinte
    0.70
     insidious
    0.68
     Ciò
    0.68
     communément
    0.65
     இறுதியில்
    0.64
     avaient
    0.64
     convi
    0.63
     dû
    0.62
     époque
    0.62
    POSITIVE LOGITS
    isoft
    0.71
    ocont
    0.70
    ース
    0.69
    astere
    0.67
    ase
    0.66
    ecause
    0.65
    abyte
    0.65
    eman
    0.64
    epro
    0.64
    امام
    0.63
    Activations Density 0.264%

    No Known Activations