© 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-2-27B
    5. 10-GEMMASCOPE-RES-131K
    7. 15184
    Prev
    Next
    INDEX
    Explanations

    Thomson and neurones

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

    This neuron activates for proper nouns, particularly company and organization names.

    oai_token-act-pair · claude-3-7-sonnet-20250219Triggered by @neilrathi

    The neuron fires on isolated “metadata-style” tokens—things like author names, acronyms, train or company codes, citation numbers, etc.—essentially single-word identifiers rather than normal prose.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_10/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
    媗
    -4.53
    姈
    -3.38
    谞
    -3.31
    蠊
    -2.81
    ᮣ
    -2.75
    诳
    -2.72
    们的
    -2.64
    潆
    -2.64
    quelize
    -2.59
    翾
    -2.59
    POSITIVE LOGITS
    y
    3.55
     -
    3.28
     
    3.25
    t
    2.86
    et
    2.81
    “
    2.73
    u
    2.72
     schwarzen
    2.66
    ’
    2.58
     I
    2.55
    Activations Density 0.016%

    No Known Activations