© 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. 22-GEMMASCOPE-RES-131K
    7. 14115
    Prev
    Next
    INDEX
    Explanations

    political, romantic, therapist, coyote, cheese, plot

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

    The neuron fires strongly on the very first prominent word (often capitalized) at the start of a new paragraph or section. It effectively detects the paragraph-initial content word.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_22/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
     of
    -1.29
    亻
    -1.01
    Mostrar
    -0.96
    例えば
    -0.96
    mbda
    -0.92
    kiksi
    -0.92
     अला
    -0.90
    ссмо
    -0.90
     ͡°)
    -0.89
     것입니다
    -0.88
    POSITIVE LOGITS
     membag
    1.23
     llamaba
    1.05
    archive
    1.05
    倒是
    1.05
    已经
    1.03
     rajut
    1.00
    Naturally
    0.99
    することは
    0.99
     Specialist
    0.98
    是一種
    0.96
    Activations Density 0.045%

    No Known Activations