© 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
    3. Gemma-2-27B
    5. 22-GEMMASCOPE-RES-131K
    7. 130972
    Prev
    Next
    INDEX
    Explanations

    The neuron activates on adverbial intensifiers—especially “well” (and “well-” compounds) and similar qualifying adverbs that boost or qualify an adjective.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408

    positive qualities after well

    np_acts-logits-general · gemini-2.5-flash-lite
    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
    を
    -2.34
    などが
    -2.33
    などを
    -2.28
    があり
    -2.19
    と
    -2.17
    が
    -2.16
    がある
    -2.13
    である
    -2.03
    がありますが
    -1.99
    などの
    -1.97
    POSITIVE LOGITS
    為に
    2.66
    Although
    2.27
    ところに
    2.11
    When
    2.09
    After
    2.00
    Even
    2.00
    Though
    1.91
    Before
    1.91
    While
    1.91
    Despite
    1.88
    Activations Density 0.022%

    No Known Activations