© 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
    3. Gemma-3-12B
    5. 24-GEMMASCOPE-2-RES-16K
    7. 7174
    Prev
    Next
    INDEX
    Explanations

    checking for "hello" or "hi"

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

    This neuron isn’t picking out any particular word type or meaning—rather it strongly activates for tokens appearing very late in the input sequence, essentially marking “far‐right” (end‐of‐context) positions.

    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
     n
    1.01
     d
    0.90
     o
    0.85
    0.81
     attraction
    0.78
     bagels
    0.78
     attractions
    0.77
     cigars
    0.76
    g
    0.75
     country
    0.74
    POSITIVE LOGITS
    vaient
    0.91
     személy
    0.85
     estratégica
    0.82
     decía
    0.81
    étais
    0.80
     sostenibilidad
    0.79
    setPreferred
    0.79
    𓏸
    0.79
     persön
    0.78
    Quieres
    0.77
    Activations Density 0.000%

    No Known Activations