© 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
    2. Gemma-3-27B-IT
    3. 31-GEMMASCOPE-2-TRANSCODER-262K
    4. 77509
    Prev
    Next
    INDEX
    Explanations

    `pmatrix` and `bmatrix`This neuron appears to activate strongly when encountering mathematical notation for matrices, specifically using `pmatrix` and `bmatrix` delimiters, which are common in LaTeX and mathematical typesetting. The presence of terms like "matrices", "$A = \begin{pmatrix} a & b \\ c & d \end{pmatrix}$", and "$A^2 B = \begin{bmatrix} ... \end{bmatrix}$" directly points to this. The tokens `[` and `{` in `MAX_ACTIVATING_TOKENS` also support this, as they are often used to define matrices or their elements.`pmatrix` and `bmatrix`

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-27b-it/transcoder_all/layer_31_width_262k_l0_small_affine
    Prompts (Dashboard)
    238,145 prompts, 512 tokens each
    Dataset (Dashboard)
    lmsys + oasst1
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     synonymous
    0.40
     tout
    0.39
     estimator
    0.39
     விர
    0.39
     estimators
    0.37
     exponencial
    0.37
    ชน์
    0.37
     interested
    0.36
     doped
    0.36
     চলচ্চিত্রের
    0.36
    POSITIVE LOGITS
    plet
    0.38
    শন
    0.38
    ırı
    0.38
    辘
    0.38
    可能です
    0.37
    명을
    0.36
    nales
    0.36
    榎
    0.36
     Bang
    0.35
    jel
    0.35
    Activations Density 0.000%

    No Known Activations