© 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. 24-GEMMASCOPE-2-RES-16K
    7. 1937
    Prev
    Next
    INDEX
    Explanations

    and questions

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

    This neuron detects the special tokens that mark interview questions or speaker‐turn labels (e.g. “Q.” and other question‐prefix markers).

    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_24_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
     pitiful
    0.65
     ALL
    0.59
     EVERY
    0.57
    ↵↵↵↵
    0.55
    <strong>
    0.55
     всі
    0.53
    的我
    0.52
     яви
    0.51
    ↵↵↵↵↵↵
    0.51
     wretched
    0.51
    POSITIVE LOGITS
     Ihrem
    0.94
     yourself
    0.88
    your
    0.86
    yourself
    0.86
     вашем
    0.84
     आपकी
    0.83
    ujete
    0.82
     Votre
    0.82
     Yourself
    0.82
     вашу
    0.82
    Activations Density 0.253%

    No Known Activations