© 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-3-27B-IT
    3. 15-GEMMASCOPE-2-TRANSCODER-262K
    4. 17664
    Prev
    Next
    INDEX
    Explanations

    The neuron appears to be detecting URLs and academic resource identifiers. It seems to latch onto words like 'org' and 'Portal', often followed by path separators or specific descriptors like 'Library' or '://', which are common in web addresses for institutions or digital resources. The other lists, while containing varied characters, don't strongly suggest a consistent alternative pattern. The 'TOP_POSITIVE_LOGITS' are mostly non-English script characters which might be related to specific encodings or diverse language use within academic contexts, but the dominant pattern points to web addresses and digital resource links.Given these observations, a concise phrase explaining the neuron's behavior would be:**urls and digital resource links**

    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_15_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
    Markdown
    0.23
     asexual
    0.22
     aleg
    0.21
    )、
    0.20
    вым
    0.20
    MOND
    0.19
    と言
    0.19
    роятно
    0.19
    としても
    0.19
     ales
    0.19
    POSITIVE LOGITS
    i
    0.25
    ième
    0.23
    𝕚
    0.23
    ią
    0.22
    ἱ
    0.22
    ي
    0.22
    ін
    0.22
     ก
    0.21
     ป
    0.21
    nya
    0.21
    Activations Density 0.001%

    No Known Activations