© 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-3-12B
    5. 24-GEMMASCOPE-2-RES-16K
    7. 7253
    Prev
    Next
    INDEX
    Explanations

    scams and attacks

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

    This neuron reliably activates on references to harmful attacks or criminal acts—e.g. theft, stealing, identity‐theft, robbery, malware exploits, and related malicious activities.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-12b-pt/resid_post/layer_24_width_16k_l0_medium
    Prompts (Dashboard)
    392,802 prompts, 256 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    pyridazine
    0.87
    荷重
    0.79
    濋
    0.78
     applicability
    0.77
    AutoSize
    0.75
    rystall
    0.73
    physiological
    0.72
    补偿
    0.72
    热爱
    0.72
    棹
    0.71
    POSITIVE LOGITS
     fraudulently
    1.98
     fraudulent
    1.95
     unscrupulous
    1.94
     nefarious
    1.91
     malicious
    1.87
     illegally
    1.84
     maliciously
    1.84
     embezzlement
    1.76
     fraude
    1.75
     deceit
    1.74
    Activations Density 2.458%

    No Known Activations