© 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. 2473
    Prev
    Next
    INDEX
    Explanations

    programming code explanations

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

    This neuron detects high-level discourse markers—the key words and phrases (often sentence-initial or heading-style tokens) that signal the start of new sections or important topic shifts.

    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
    পান
    0.59
     bahasa
    0.54
     socios
    0.54
    <{},
    0.54
     kalah
    0.53
    喫
    0.52
     berkel
    0.52
    Ciudad
    0.51
     bawa
    0.51
    דיה
    0.51
    POSITIVE LOGITS
     被
    0.60
     DOCUMENT
    0.58
     れ
    0.57
     monitored
    0.56
     Document
    0.55
    *
    0.55
     document
    0.55
    根
    0.54
    Op
    0.54
     分
    0.54
    Activations Density 0.009%

    No Known Activations