© 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. 112380
    Prev
    Next
    INDEX
    Explanations

    verbs that take 'to' or 'do'

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

    The neuron is picking out very high-frequency function words (common stopwords like we, you, have, can, do, made, live, etc.).

    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
    }$
    -1.98
    ↵↵
    -1.94
     Because
    -1.63
    e
    -1.44
    のが
    -1.41
     in
    -1.40
     does
    -1.34
     t
    -1.34
     about
    -1.31
     Beſ
    -1.31
    POSITIVE LOGITS
     onze
    1.75
     semelh
    1.72
    퐿
    1.70
     such
    1.66
     ráp
    1.65
    傒
    1.64
    ᱙
    1.63
     huit
    1.59
    珰
    1.59
    旄
    1.58
    Activations Density 0.104%

    No Known Activations