© 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
    3. Gemma-2-27B
    5. 34-GEMMASCOPE-RES-131K
    7. 1252
    Prev
    Next
    INDEX
    Explanations

    customer service or experts answering

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

    The neuron is detecting lines that announce or label the speaker—e.g. self-introductions (“Hi, I’m Dr. Jackie…”) and metadata headers like “JACUSTOMER-29g2cvk7:”

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_34/width_131k
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     دریافت‌شده
    -0.87
    }}(\
    -0.72
    Ald
    -0.71
     Ordner
    -0.71
    mdp
    -0.69
     chiến
    -0.68
    维护
    -0.68
     cmb
    -0.67
    .—
    -0.67
     monkeys
    -0.67
    POSITIVE LOGITS
    Customer
    1.33
     JA
    1.25
    answer
    1.07
    customer
    1.04
     customer
    1.00
     experts
    0.99
    JA
    0.98
    Experts
    0.98
     Experts
    0.98
     answer
    0.96
    Activations Density 0.020%

    No Known Activations