© 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
    2. Gemma-3-12B
    3. 31-GEMMASCOPE-2-RES-16K
    4. 2653
    Prev
    Next
    INDEX
    Explanations

    However, these

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

    The neuron fires most strongly on sentence-initial discourse cues—introductory adverbs and phrases (e.g. “In this case,” “According to,” “Traditionally,” “Originally,” etc.) that mark topic shifts or new sections.

    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
    but
    0.72
    But
    0.56
     but
    0.55
    BUT
    0.53
     Nhưng
    0.52
    だけど
    0.50
     けど
    0.49
     Или
    0.49
     அல்லது
    0.49
     αλλά
    0.49
    POSITIVE LOGITS
     there
    0.63
     researchers
    0.57
     however
    0.56
     অবশ্য
    0.55
     various
    0.53
     variations
    0.53
     policymakers
    0.53
     several
    0.52
     dunque
    0.52
     heterogeneity
    0.51
    Activations Density 0.186%

    No Known Activations