© 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. 38191
    Prev
    Next
    INDEX
    Explanations

    how could

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

    The neuron is a “could” detector – it spikes on the modal auxiliary “could” (and related possibility/ability contexts, often in questions).

    oai_token-act-pair · o4-miniTriggered by @jyhe0408

    rhetorical questions expressing disbelief or incredulity, particularly those beginning with "how could" or similar constructions.

    oai_token-act-pair · claude-4-5-sonnetTriggered by @jyhe0408

    incredulous modal constructions expressing possibility or ability, especially “how”-questions built around a modal of possibility.

    oai_token-act-pair · gpt-5Triggered 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
     ?></
    -0.90
    rícula
    -0.85
    ächtig
    -0.84
     bepaalde
    -0.83
     philipp
    -0.83
     Ausdrücke
    -0.82
    ΟΥ
    -0.80
     toti
    -0.79
    änden
    -0.78
     afp
    -0.77
    POSITIVE LOGITS
     possibly
    3.64
     possible
    3.19
     Possibly
    2.77
    possibly
    2.56
    Possibly
    2.52
    possible
    2.22
    Possible
    2.22
     posiblemente
    2.19
     Possible
    2.13
     возможно
    1.91
    Activations Density 0.041%

    No Known Activations