© 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-3-27B-IT
    5. 40-GEMMASCOPE-2-RES-262K
    7. 2056
    Prev
    Next
    INDEX
    Explanations

    refusal to assist

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

    Instances where the text refers to the assistant itself (the model/agent identity token).

    oai_token-act-pair · gpt-5-miniTriggered by @vetterc0
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-27b-it/resid_post/layer_40_width_262k_l0_medium
    Prompts (Dashboard)
    238,145 prompts, 512 tokens each
    Dataset (Dashboard)
    lmsys + oasst1
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     Columns
    0.45
     attribution
    0.42
     columns
    0.42
     ColumnName
    0.41
     column
    0.38
     rights
    0.37
    trait
    0.37
     #
    0.37
     assimilation
    0.37
     integration
    0.36
    POSITIVE LOGITS
    >′
    0.42
    Rew
    0.39
     المستقيم
    0.38
    баев
    0.38
    स्करी
    0.38
    Negative
    0.38
    Sorry
    0.38
    бари
    0.37
    फेस
    0.37
    Android
    0.37
    Activations Density 0.003%

    No Known Activations