© 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 EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-3-12B
    5. 31-GEMMASCOPE-2-RES-16K
    7. 897
    Prev
    Next
    INDEX
    Explanations

    The neuron strongly activates on blocks of direct speech—i.e. quoted sentences (often surrounded by quotation marks and speaker attributions).

    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
    私の
    0.71
    我的
    0.60
    在他的
    0.58
    His
    0.57
    HIS
    0.55
    彼の
    0.55
    There
    0.54
    给我
    0.54
    তাঁর
    0.54
    給我
    0.53
    POSITIVE LOGITS
     she
    1.09
     he
    1.00
     она
    0.86
     वह
    0.76
     তিনি
    0.76
     President
    0.75
     він
    0.75
     вона
    0.75
     Başkan
    0.75
     DeSantis
    0.73
    Activations Density 0.758%

    No Known Activations