© 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-2-27B
    3. 22-GEMMASCOPE-RES-131K
    4. 54363
    Prev
    Next
    INDEX
    Explanations

    code tokens and punctuation

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

    The neuron is picking out non‐natural-language “markup” tokens—things like HTML entities (e.g.  ), punctuation‐heavy link or code fragments, URLs, and uppercase acronyms.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_22/width_131k
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     of
    -2.39
     it
    -1.88
     on
    -1.57
     had
    -1.54
     for
    -1.52
     with
    -1.36
     />';
    -1.34
     out
    -1.33
    :
    -1.28
     while
    -1.27
    POSITIVE LOGITS
    ꟷ
    1.46
     ruines
    1.45
    othes
    1.41
    があり
    1.38
    があるので
    1.36
    以及
    1.34
     quelques
    1.34
    还有
    1.33
    などが
    1.33
     firs
    1.29
    Activations Density 0.034%

    No Known Activations