© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-27B-IT
    3. 40-GEMMASCOPE-2-RES-262K
    4. 3900
    Prev
    Next
    INDEX
    Explanations

    Human, Assistant, agent

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

    This neuron detects speaker/role labels and dialogue turn markers (tokens that indicate who is speaking, like "assistant", "user"/"human", character names, or bracketed turn tags).

    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
    りの
    0.38
     HPLC
    0.34
    々に
    0.33
     permeable
    0.33
    %";
    0.32
     dilat
    0.32
     psychosocial
    0.31
     DBMS
    0.31
     einer
    0.31
     ___
    0.30
    POSITIVE LOGITS
    Actually
    0.35
    uddho
    0.33
    Okay
    0.32
     LikeLike
    0.32
     తప్ప
    0.32
    这边
    0.32
     Okay
    0.31
     যাক
    0.31
     โอ้
    0.31
     Yeah
    0.30
    Activations Density 0.015%

    No Known Activations