© 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. 73587
    Prev
    Next
    INDEX
    Explanations

    wrongdoing and abuses

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

    This neuron detects terms referring to wrongful acts or abuses—particularly crimes, violations, or human-rights abuses.

    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
    驺
    -1.52
     nabízí
    -1.52
     profundos
    -1.52
    '
    -1.52
     ignited
    -1.50
    椅
    -1.50
    inqu
    -1.49
     ofreciendo
    -1.48
     theres
    -1.47
    emplares
    -1.47
    POSITIVE LOGITS
    7
    2.17
    s
    2.16
    4
    1.98
    8
    1.98
    0
    1.74
    ta
    1.73
    </h4>
    1.63
    美味しそう
    1.63
     dugaan
    1.58
    _
    1.55
    Activations Density 0.010%

    No Known Activations