© 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 EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-27B-IT
    3. 27-GEMMASCOPE-2-TRANSCODER-262K
    4. 196141
    Prev
    Next
    INDEX
    Explanations

    This neuron likely activates around the concept of 'Ethical Considerations' or 'Ethical approval', particularly in contexts related to reviews, approvals, or obtaining consent. The presence of "Eth", "ethical", "Ethics", "approval", and "considerations" in `MAX_ACTIVATING_TOKENS` strongly suggests this. The subsequent tokens like "number", "considerations", and "review" in `TOKENS_AFTER_MAX_ACTIVATING_TOKEN` further reinforce this. The `TOP_POSITIVE_LOGITS` listing words like "cancelar" (to cancel) and "canceling" might indicate a role in decisions or processes involving ethical clearance or cancellation based on ethical grounds, though this is less pronounced than the core ethical theme.Therefore, a concise explanation would be: **ethical considerations and approval**ethical considerations and approval

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-27b-it/transcoder_all/layer_27_width_262k_l0_small_affine
    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
     couples
    0.44
     individuals
    0.42
     Ny
    0.42
     N
    0.39
     specimens
    0.39
     specimen
    0.39
    Ny
    0.39
     species
    0.38
     pairs
    0.38
    +
    0.38
    POSITIVE LOGITS
     cancelar
    0.46
     finalList
    0.45
    օ
    0.45
    旮
    0.44
     autorización
    0.43
    闼
    0.42
    尣
    0.42
     professeurs
    0.41
    вающих
    0.41
     canceling
    0.40
    Activations Density 0.000%

    No Known Activations