© 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. 54005
    Prev
    Next
    INDEX
    Explanations

    don't do or advise against

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

    The neuron flags advisory or warning language—phrases that tell you “don’t,” “shouldn’t,” “avoid,” “advise,” “recommend,” 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_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
     morreu
    -1.16
     všechny
    -1.11
    บ้าง
    -1.09
     instead
    -1.09
    CAPT
    -1.07
    quée
    -1.04
     encontrou
    -1.02
    낍
    -1.01
    quiao
    -1.01
    CAPÍTULO
    -1.01
    POSITIVE LOGITS
     any
    1.43
     unless
    1.30
     terlalu
    1.29
     EVER
    1.21
     anymore
    1.20
     use
    1.14
     it
    1.13
     anyone
    1.09
     for
    1.05
    有任何
    1.05
    Activations Density 0.034%

    No Known Activations