© 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-3-12B
    5. 31-GEMMASCOPE-2-RES-16K
    7. 13406
    Prev
    Next
    INDEX
    Explanations

    malware, malicious, Malala

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

    This neuron activates strongly on the subword “Mal,” i.e. on occurrences of “Mal” within words (especially proper names like Malm ö, Malware, Malignant, Maldon, etc.).

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

    the text fragment "Malm" or similar word stems, particularly in names and place references (like Malmö, Malware, Malignant, Malick, Malm, Malavé, Malise, Malad, Malen, mal

    oai_token-act-pair · claude-4-5-sonnetTriggered by @jyhe0408

    words containing the “mal” morpheme, especially at the start of names or terms.

    oai_token-act-pair · gpt-5Triggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-12b-pt/resid_post/layer_31_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
     Plains
    0.56
     plains
    0.51
     afterPath
    0.50
     कस
    0.50
    才是
    0.49
    adeus
    0.49
    Notable
    0.48
    驭
    0.48
    PH
    0.47
    Dur
    0.47
    POSITIVE LOGITS
     mal
    0.77
     MAL
    0.75
    Mal
    0.69
     Mal
    0.68
    mal
    0.67
     مال
    0.64
    achite
    0.63
     afore
    0.61
    वाहक
    0.61
    MAL
    0.59
    Activations Density 0.075%

    No Known Activations