© 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. 5565
    Prev
    Next
    INDEX
    Explanations

    This neuron activates strongly on the framing and attribution language used in news‐wire or press‐release style reporting—things like sentence‐initial gerunds that summarize the story (“Highlighting…”), reporting verbs (“said,” “noted,” “informed”), and the nearby punctuation that marks those 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
    umably
    0.75
    噱
    0.71
     интересу
    0.69
     asshole
    0.69
     ആവശ
    0.65
     ঝামে
    0.65
    讨厌
    0.64
     suspected
    0.63
     psik
    0.63
     нужен
    0.63
    POSITIVE LOGITS
     achievements
    2.19
     successes
    2.09
     accomplishments
    2.02
     Achievements
    1.92
     logros
    1.89
    Achievements
    1.79
    成功
    1.76
     başarı
    1.76
     সাফল্যের
    1.66
     সাফল্য
    1.62
    Activations Density 0.552%

    No Known Activations