© 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. 24-GEMMASCOPE-2-RES-16K
    4. 860
    Prev
    Next
    INDEX
    Explanations

    This neuron isn’t detecting any particular word or concept in the text but rather its location: it fires strongly only at tokens that occur at a specific position (i.e. it’s a “positional” neuron) rather than because of their identity or meaning.

    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
    ()!=
    0.72
    猾
    0.71
    悟
    0.70
     কথায়
    0.69
     เหรอ
    0.69
     sprang
    0.67
    Ꭾ
    0.67
     klubu
    0.66
    Թ
    0.66
    訨
    0.66
    POSITIVE LOGITS
     TOKEN
    0.81
     archivo
    0.74
     Algunos
    0.71
     viviendas
    0.70
     incumpl
    0.70
     прекра
    0.68
     piezas
    0.67
    TOKEN
    0.66
     ofrecemos
    0.66
     Uml
    0.66
    Activations Density 5.875%

    No Known Activations