© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it
    Neuronpedia
    Jacobian LensNEW
    Natural Language
    Autoencoders
    MAY 26
    Assistant AxisJAN 26Circuit TracerAPR 26SteerSAE EvalsExportsinterp-engineNEW Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-12B
    3. 31-GEMMASCOPE-2-RES-16K
    4. 3369
    Prev
    Next
    INDEX
    Explanations

    The neuron strongly activates on uppercase‐heavy, technical or code‐style tokens—e.g. error messages, acronyms, all-caps fragments, or similarly formatted “computer-ese.”

    oai_token-act-pair · o4-mini
    Triggered 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
    Negative Logits
    1
    0.81
    -
    0.75
    2
    0.74
    A
    0.74
    T
    0.73
    E
    0.72
    C
    0.71
    N
    0.70
    G
    0.70
    3
    0.70
    POSITIVE LOGITS
     えっと
    0.89
     可愛
    0.88
    ॓
    0.85
    க்ஸாண்ட
    0.84
     هایی
    0.83
     ましょう
    0.82
     های
    0.81
     感じ
    0.80
     入っ
    0.80
    ächer
    0.77
    Activations Density 0.624%

    No Known Activations