© 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
    2. Gemma-3-12B
    3. 24-GEMMASCOPE-2-RES-16K
    4. 9582
    Prev
    Next
    INDEX
    Explanations

    uncontrolled

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

    The neuron is essentially a “position‐counter” that fires strongest right at the start of a new text chunk and then its activation steadily decays the further you move away from the beginning. In other words, it’s detecting “how close am I to the start of the document?”

    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
    i
    0.73
    got
    0.72
    bés
    0.71
    িস
    0.70
    aligned
    0.69
     imposs
    0.69
     समर्
    0.69
    یاں
    0.68
    ધા
    0.68
     இடையே
    0.67
    POSITIVE LOGITS
     alcanzar
    0.90
     rectángulo
    0.90
     geval
    0.85
     secuencia
    0.85
    тэй
    0.84
     krok
    0.82
     zelfs
    0.81
     thấm
    0.80
     cumplir
    0.80
     botella
    0.80
    Activations Density 0.000%

    No Known Activations