© 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-2-27B
    5. 22-GEMMASCOPE-RES-131K
    7. 33785
    Prev
    Next
    INDEX
    Explanations

    foreign words or word stems

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

    The neuron strongly activates on subword fragments that are the initial piece of a longer, split‐up word (i.e. the first BPE or WordPiece token of a multi‐token word).

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_22/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
    When
    -3.55
    Despite
    -3.36
    Because
    -3.28
    Before
    -3.23
    During
    -3.17
    After
    -3.17
    Taking
    -3.17
    Perhaps
    -3.14
    If
    -3.14
    What
    -3.14
    POSITIVE LOGITS
     .......
    2.98
     ........
    2.94
     ..........
    2.89
    ᕮ
    2.89
     .........
    2.81
     Dinas
    2.77
     soooo
    2.72
    蓠
    2.69
     "*
    2.67
     …..
    2.64
    Activations Density 0.111%

    No Known Activations