© 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
    2. Gemma-3-4B-IT
    3. 22-GEMMASCOPE-2-RES-65K
    4. 5569
    Prev
    Next
    INDEX
    Explanations

    names preceded by bold markers*Self-correction: The prompt asks for something specific without mentioning tokens or patterns. The list shows `**` in MAX_ACTIVATING_TOKENS aligning with names in TOP_ACTIVATING_TEXTS and TOP_POSITIVE_LOGITS. The TOKENS_AFTER_MAX_ACTIVATING_TOKEN are also names/initials. The explanation should capture the detected entity type. The `**` token often acts as a marker for bolded words, and bolded words in this context are typically names. Therefore, the neuron seems to focus on names that are presented in a specific format (bolded).*names in bold

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-4b-it/resid_post/layer_22_width_65k_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
     Barca
    0.82
     Minaj
    0.73
     দলটি
    0.72
     जानवर
    0.71
     FastAPI
    0.70
     az
    0.69
    公主
    0.69
    ާތ
    0.69
    **
    0.68
    (**
    0.68
    POSITIVE LOGITS
     Sean
    1.10
     Steve
    1.04
    Steve
    1.04
     Alan
    1.02
     Mike
    1.02
     Matthew
    1.01
     Andrew
    1.01
     Michael
    0.98
     William
    0.97
     Josh
    0.97
    Activations Density 0.076%

    No Known Activations