© 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. 531
    Prev
    Next
    INDEX
    Explanations

    learned

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

    The neuron lights up on semantically “contentful” words—nouns, verbs, and adjectives that carry real meaning—while largely ignoring function words.

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

    past participle verbs, particularly in contexts describing completed actions or experiences.

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

    action and process language—primarily verbs and verbal forms (often -ing or participles) that describe doing, moving, seeing, using, communicating, or causing outcomes.

    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
    ของการ
    0.35
    altura
    0.31
    ของ
    0.30
    将
    0.29
    اگر
    0.28
    archivo
    0.28
    displaystyle
    0.27
    ما
    0.27
    如果
    0.27
     जाएगी
    0.26
    POSITIVE LOGITS
     out
    0.34
    在一起
    0.33
     elsewhere
    0.32
     în
    0.31
    in
    0.30
     abroad
    0.30
     WITH
    0.29
     into
    0.29
     consciously
    0.29
     outside
    0.28
    Activations Density 9.955%

    No Known Activations