© 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
    3. Gemma-2-27B
    5. 34-GEMMASCOPE-RES-131K
    7. 89906
    Prev
    Next
    INDEX
    Explanations

    the right thing

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

    The neuron is a “function‐word” detector: it strongly activates on high‐frequency closed-class tokens (e.g. the, did, does, made, it, work, acronyms like CMA) rather than on content words.

    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
     新作
    -0.94
    Продол
    -0.93
    Gui
    -0.90
    importance
    -0.90
     severa
    -0.89
     impecable
    -0.88
    섹
    -0.88
     laajenta
    -0.87
     évé
    -0.86
     poesía
    -0.85
    POSITIVE LOGITS
     same
    2.03
     necessary
    2.02
     requisite
    1.42
     required
    1.38
     right
    1.37
     equivalent
    1.33
     appropriate
    1.30
     needed
    1.20
     impossible
    1.15
     usual
    1.15
    Activations Density 0.016%

    No Known Activations