© 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. 22-GEMMASCOPE-RES-131K
    7. 98285
    Prev
    Next
    INDEX
    Explanations

    not easy, same talent, paradigm shift

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

    The main thing this neuron does is pick out the core predicate words of a sentence—especially auxiliaries and negations and their complement adjectives or verbs (e.g. "is not easy," "won’t budge," "is not some dummy").

    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
     masina
    -1.02
     Jeffery
    -1.00
     GETTING
    -0.99
     plais
    -0.98
     Arquitectura
    -0.97
     menant
    -0.97
     vedette
    -0.96
     Senin
    -0.96
     Clínica
    -0.94
    certainly
    -0.94
    POSITIVE LOGITS
     only
    1.05
     itself
    1.01
     ironically
    0.93
     similarly
    0.93
    היו
    0.93
     any
    0.90
     stessi
    0.89
     hornear
    0.89
     singing
    0.85
     some
    0.85
    Activations Density 0.059%

    No Known Activations