© 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. 31-GEMMASCOPE-2-RES-16K
    7. 3061
    Prev
    Next
    INDEX
    Explanations

    The neuron strongly activates on numeric date tokens—especially multi‐digit years (e.g. “1491,” “1470,” etc.)—marking explicit historical-year mentions.

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

    historical buildings, religious structures, and architectural landmarks.

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

    mentions of notable architecture, especially historic or religious buildings and monuments, including their structures and construction details.

    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_31_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
     невро
    0.73
     उबाल
    0.70
     onboarding
    0.68
     covariates
    0.68
     probation
    0.68
     Probation
    0.68
     معاش
    0.67
     interpersonal
    0.66
     hábitos
    0.66
     infek
    0.65
    POSITIVE LOGITS
    建築
    2.20
    建筑
    2.13
     건축
    2.03
     buildings
    1.92
    buildings
    1.87
     архитек
    1.85
     architectural
    1.82
    architect
    1.82
    architecture
    1.77
     edificios
    1.77
    Activations Density 0.706%

    No Known Activations