© 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. 818
    Prev
    Next
    INDEX
    Explanations

    speaking softly, exhaling

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

    The neuron detects narrative “speech‐tag” verbs and adjoining character action verbs that frame or report dialogue (e.g. said, asserted, insisted, looked, grinned).

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

    words indicating reported speech, particularly "said" and related dialogue attribution verbs.

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

    dialogue tags and nearby action/gesture verbs that attribute or accompany spoken lines.

    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_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
    0.85
     f
    0.60
     (
    0.60
    -
    0.60
     l
    0.58
     D
    0.55
     F
    0.54
     V
    0.54
     S
    0.54
     '
    0.54
    POSITIVE LOGITS
     sambil
    0.84
     excitedly
    0.84
    óso
    0.81
     angrily
    0.77
     reluctantly
    0.72
    ྛ
    0.71
     aloud
    0.71
     muttered
    0.71
     председа
    0.71
    ‌.
    0.70
    Activations Density 0.032%

    No Known Activations