© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW
    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-12B
    3. 24-GEMMASCOPE-2-RES-16K
    4. 11047
    Prev
    Next
    INDEX
    Explanations

    other examples of

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

    The neuron fires on high-frequency function words and other common short tokens (e.g. people, who, to, are, a), effectively marking connective and pronoun‐type words.

    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
    becomes
    0.70
    ATION
    0.67
    物体
    0.67
    巧妙
    0.67
    约为
    0.65
    亰
    0.65
    world
    0.64
    reducing
    0.64
     и
    0.63
    神器
    0.63
    POSITIVE LOGITS
     outspoken
    0.92
     člen
    0.88
     ehemal
    0.87
     Geschäftsführer
    0.80
     handicapped
    0.80
     nominated
    0.79
     Mitglied
    0.79
     participated
    0.79
     bereits
    0.79
     handicap
    0.79
    Activations Density 0.417%

    No Known Activations